Dvoretsky's Endgame Manual

l ... �e5?!
A strange move indeed! Common sense
Marin - Slovineanu would indicate Black should exchange on h4.
Romania 1999 The side that is down material should usually
exchange pawns whenever possible.
Cifuentes believes that Black would stand
poorly then, and demonstrates this with the fol­
5-16 lowing variation: l . . .gh 2 gh Ab8 3 �h3 Aa7 4
�g4 �e5 5 �h5 0 �d5 6 Af5! (6 �xh6? would
be premature: 6 . . . Axe3+ 7 �g7 Aa7! 8 Axe4+
�xe4 9 h5 �d5 10 h6 �c6=) 6 . . . �e5 7 �g6
w �d5 (7 . . . h5 8 Ah3 �d6 9 Ag2 �e5 1 0 �xh5
�f5 1 1 Axe4+!) 8 h5 �e5 9 Axe4! �xe4 1 0
�xh6 Axe3+ 1 1 �g6+- .
Cifuentes' analysis is completely uncon­
The game ended very quickly: 1 �c4?!
vincing. Why should Black allow White's king
.Q.e7 2 �b5 �g7 3 .Q.d3 h6 Drawn.
to attack his h-pawn? For example, he could try
Marin, in Informant 75, gave his 1 �c4 two
3 . . . h5!? 4 Axh5 �c6. It would be much simpler,
question marks. He assessed his position as win­
however, to set up an impregnable fortress by
ning, and demonstrated this with the following
giving up Black's main weakness - the e4-pawn
variation:
1 �c6 Ae7 2 b5 .llc 5 3 �d7 Ab4 4 e6 Ac5 - at once.
5 Ad3 Ae7 6 ..lle4 .llc 5 7 f5 gf 8 .llxf5 !iJ..e7 9 Let's continue: 3 . . . �e5 4 �g4 �f6! 5 !iJ..xe4
�c6! (D. Rogozenko) - apparently it was this Ac7. Now the h-pawn is untouchable - 6 �h5 is
last move, later pointed out by his colleague, that met by 6 . . . �g7. White has to bring his king to
the GM failed to notice during the game - the queenside; but the most he can achieve there
9 . . . Axg5 (9 . . .!iJ..d8 10 Axh7) 10 �xb6 �e7 1 1 is the win of the bishop for his b- and e-pawns.
�c6 Af4 12 b6 h5 1 3 �b7 Ae3 14 �c7+- . But then Black's king goes to h8, with an elemen­
Evidently, neither Marin nor Rogozenko was tary draw (the enemy bishop does not control
aware of the Berger-Kotlerman endgame. Other­ the rook pawn's queening square). And this im­
wise, they would clearly have seen that 1 2 . . .�d8! portant defensive resource comes about precisely
(instead of12 . . .h5??) would secure Black the draw. because of the exchange of pawns at h4.
Actually, if he wishes, Black could even keep his Even in Cifuentes' line 3 . . . Aa7 4 �g4 �e5
h-pawn (which, in fact, has not the slightest value 5 �h5, it's still not too late to return to the right
anyway) by playing 1 1 . . .h5 (instead of 1 l . . .Af4) plan: 5 . . . �f6!, since after 6 �xh6 (6 Axe4 �g7=
1 2 b6 (12 �c7 Af4+ 1 3 �c8 Ae3) 12 . . . �d8 1 3 ) 6 . . . .\lxe3+ 7 �h7 Af4 8 Axe4 Ab8, White is
�b7 Ae3!=, or 1 3 b7 ..llf4 1 4 �b6 Ab8!=. unable to queen the h-pawn: 9 Af3 Ac7 10 h5
�g5 1 1 h6 !iJ..e5 12 Ae2 Ab8 13 �g7 Ae5+.
Cifuentes - Langeweg Black's refusal to trade pawns probably
El Vendrell l 996 stems from the fact that Langeweg did not want
to free the g3-square for White's king. The king
cannot approach through the h3-square, which
can be seen from the line 2 �h3 �d6! 3 �g4
5-1 7 �c7 (3 . . . gh) 4 h5 (4 hg hg 5 Axe4 Axe3=)
4 . . . Axe3 (4 . . . �xb7 5 Axe4+ �c7=) 5 Axe4
Ad2 6 �f5 g4! =.
2 h5!? �d5?
B? This was, evidently, the decisive error! As
Bologan pointed out, Black had a simple draw
with 2 . . . g4! followed by . . . Ab8. Black's king
easily defends the kingside pawns (3 .ll e 8

1 00
 Opposite-Colored Bishops

�f5); and the g3-pawn will drop as soon as Exercises

the white king leaves its side.
3g41
After fixing the kingside, White can now
direct his king to the opposite side of the board,
restrict his opponent's movements, and finally 5-19
break through the center to reach the weak pawn
at h6.
3...�e54�f2 .Q.b8 5�e2 .Q.a76�d2 5/3
�d5 7 �c30 .Q.b8 8 .Q.f7+ �c5 9 .Q.g6 B?
�d5 10 �b4! .Q.g3 11 �b5 .Q.c7 12 �a6
.Q.b8 13�b60 �e514 �c6 �e6
Black had to give up the e4-pawn anyway
(because of the mortal threat of �d7-c8 ), but in
a far less favorable situation.
15 .Q.xe4 .Q.g3 16 .Q.f5+ �e7 17 �b6
.Q.b8 18 e4 �d6
5-20

5/4
5-18
W?

W?

19 e5+!�e7
19...�xe5 loses to 20 'i!?ic6; and if 19 ...�d5,
then 20 Ac8 'i!?txe5 21 'i!?tc6 �f6 2 2 'i!?td7 'i!?if7 2 3
�dB+-. Now imagine the same position, but
without the g-pawns: Black could then simply
capture on e5.
20 .Q.c2 �e621 .Q.b3+�e722 .Q.a2 0
(if 2 2 �c6?? AxeS 23 'i!?td5 �f6= ) 22...�d723
�c51 .Q.xe5 24 �d5 .Q.f4 25 �e4 �e7 26
�f5 .Q.c7 27�g6 Black resigned.

101
 Dvoretsky's Endgame Manual

The King Blockades the Passed Pawn

Quite often the stronger side will have a White's king goes to a5, to free his bishop from
passed pawn, which needs to be blockaded by the defense of the b6-pawn. The f6-pawn will
either the king or the bishop. then have to advance, and White's king will re­
Thefirst defensive system: The king block­ turn to the kingside.
ades the enemy passed pawn, while the bishop
defends its own pawns. This is the basic and usu­ Kotov- Botvinnik
ally the most secure defensive arrangement. USSR eh, Moscow 1955
Attempts to break down the first defensive
system always involve the creation of a second
passed pawn, frequently by means of a pawn
breakthrough. 5-22

J. Speelman

B?

5-21

A classic example of the destruction of the

first defensive system. The decisive breakthrough
w aims to create a second passed pawn.
l ...g51 1
A mistake would be l...'it'g4? 2 Ae7=, and
if 2 ...'it'f3 (this position occurred in the game:
In such situations the bishop can easily White played Ac5?), then 3 rtid2! b2 4 rtlc2 rtixe3
handle the defense of the kingside, so a drawn 5 'it'xb2 'tlxf4 6 'tlc3=.
outcome should come as no surprise. 2 fg
1 f5 (the only try) l gf?!
... Hopeless is 2 hg h4 3 f5 (3 Ad6 Af5 4 g6
This move makes Black's task a bit more itxg6 5 f5 itxf5 6 'tlxb3 'it'g2) 3 ...Jlxf5 4 'it'xb3
difficult. I..jid3!= is safer. h3 5 Ad6 rtlxe3.
2 \tlxf5 Ae6+! 3\tlg5 Ag4 = 2 ...d4 +1
The assessment of the position would The b3-pawn must be defended 2 .. .'<t'g3? 3
change if Black incautiously played 2...rtic6? (in­ 'tlxb3 =.
stead of 2.. . ite6+!): 3ed
3 rtlg5 i;te2 4 h3! (but not 4 rtif6 Ac4 5 h3?, An interesting sideline is 3 Axd4 rtig3 4 g6
in view of 5... itfl! 6 g4 h4! 7 'tlxf7 Axh3 8 g5 «t'xh4 5 «t'd2 «t'h3! (5.. .'<t'g3 6 Ae5+ «t'g2 7 Af6)
ilfl - the advance of the h-pawn distracts the 6 Af6 (6 'it'e2 «t'g2! 7 Af6 h4) 6... h4 (threaten­
bishop from the defense of the b6-pawn). ing 7..'it'g3) 7 'it'e2 «t'g2-+ ("pants").
4...rtib7 (4 ....QJ1 5 g4 hg 6 h4+-) 5 Ad4 3... \tlg3
'tlc6 6 g4 hg 7 h4+-. The careless 3 ...rtig4? would have led to a
White has achieved his aim: the creation of draw after 4 d5 Axd5 5 Af2.
a second passed pawn! 4 Aa3
7...g3 8 'it'f4 (8 h5 f6+ 9 rtig6?! is much less Note the black bishop's excellent position
convincing: 9.. .f5 10 h6 f4 11 rtlg5 f3 12 h7 f2- in the variation 4 Ae7 «t'xh4 5 g6+ «t'g4. It pro­
White's play might be strengthened, however, by tects the b3-pawn and restrains both enemy
9 'it'h4!) 8 .. .ith5 (8 ...g2 9 'tig5 /::,. 10 h5) 9 'tlxg3 pawns along the single diagonal a2-g8. White
f6 10 rtlf4 Ag6 11 rtlg4 rtlb7 12 h5 Ah7. has no counterplay, so Black just advances his
Black has set up a barrier, but one which h-pawn and wins the bishop for it.
can be overcome without much difficulty. 4... \tlxh4 5 \tld3 \tlxg5

102
 Opposite-Colored Bishops

Another strong line is 5...<it'g3 6 'it'e4 h4 7

d5 h3 8 de h2 9 i.td6+ 'it'g4 10 Axh2 b2 11 e7
b1�+.
6<it'e4h47<it'f3(7 d5 Axd5+) 7 Ad5+ .•• 5-24
White resigned. After 8 'it'f2, Black's king
goes after the b3-pawn. The bishop, meanwhile,
defends the h-pawn, while restraining the d-pawn
along the diagonal c8-h3. B

Tr-aulc()medles

Bell on- Minic

13...'it'f4!
Siegen ol 1970
Zugzwang! White's bishop is torn apart: on
the one diagonal, it protects the g5-pawn; on the
other, it controls the aS-square. On 14 i.tc7+
<it'xg5, Black moves his king to d3 and plays ..g6-
5-23
.

g5-g4-g3+, when Axg3 loses to . a6-a5, and

. .

'it'xg3 to ... <it'e2.

After 14 Ae7 a5! 15 ba b4, the king goes to
the queenside once again, to win the bishop for
W?
the b-pawn. White has no counterplay, since the
black bishop does everything on the one diago­
nal a8-h l , defending the f3-pawn and stopping
This was the adjourned position, in which both enemy passers.
White sealed his 41st move. After analyzing in The king retreat is no help either.
their rooms, the players agreed to a draw with­ 14 'it'fl 'it'e3 15 Ac7 a5! 16 Axa5 (16 ba
out resumption. Black's positional advantage ap­ b4) 16...Ad5, followed by 17...Ac4+ and 18...f2+.
peared insufficient for victory to Minic. Judg­ A fter giving some thought to the final posi­
ing from his comments in the Informant, he was tion of this variation, we come to understand that
convinced by the following line: 1 Ad8 Ac6 2 White's own pawn at c5 is in his way, because it
i.tc7 'it'f5 3 i.td8 'it'g4 4 Ae7 a5 5 ba <it'f5 6 a6 blocks the important a7-g l diagonal. So White
'it'e4 7 a7 <it'd4 8 a8� i.txa8 9 c6 Axc6 10 i.td6, must rid himself of it.
when the a3-pawn is securely protected, and the 1 c6!!
draw is obvious. The only saving line. In fact, Bellon prob­
It's surprising that even after home analy­ ably sealed the other move instead. Otherwise,
sis, neither the players themselves nor their team­ after the game ended, this line would have been
mates were able to solve this rather simple posi­ revealed in the annotations.
tion. In point of fact, its evaluation hinges on 1 •.• Axc62 Ads <it'd33Ac71
the sealed move. "Pawn in the crosshairs"- it's important to
After 1 i.td8? Ac6!, Black wins. To begin force it to move onto the same color square as its
with, he must simply capture the a3-pawn (since bishop.
the bishop cannot protect it), and then the threat 3...f3 4 Ads <it'c2 5 Ac7 <it'b36 Ads
of the ... a6-a5 breakthrough will become more <it'xa37Aa5 <it'b3S <it'e3<it'c49 <it'f2 <it'd3
serious. Taking the pawn at a5 would give Black 10 Ac7 <it'e411 Ab6Ad512 Ac7<it'f513
his second passed pawn. AdS
2 i.tc7 f3 3 Ad8 'it'd3 4 Ac7 'it'c2 5 Ad8 This is the same position as in the last dia­
�b3 6 'it'e3 'it'xa3 7 Aa5 'it'a2! 0 8 'it'd2 (8.'it'd3 gram - except that there is no pawn at c5. Here
�b3) 8...<it'b2 0 9 'it'e3 'it'c3 0 . Having done its Black gets nothing from 13...a5 14 Axa5 'it'xg5,
job, the king returns to the other side. since the connected passed pawns are easily
10 'it'f2 'it'd3 11 Ad8 'it'e4 12 Ac7 'it'f5 13 blockaded on the dark squares. If White had not

103
 Dvoretsky's Endgame Manual

forced the timely advance of the f-pawn, with the

pawn at f4 this position would be lost, of course.
13 �f414�fll
•.•

5-26
Now on 14... �e3, White has 15 Ab6+ -
this check was the reason behind the pawn sac­
ri fice.
14... Jl,c4+ 15 �f2
W?
On 15... Ad3, White cannot play either 16
'i!te1? �g3! 17 Ac7+ �g2 or 16 �g1? 'i!te3! 17
Ab6+ 'i!te2, followed by 18... a5. However, he
does have 16 Ac7+, exploiting the fact that the
White might still have saved the game by
f3-pawn is not protected by the bishop
playing 7 'i!td2! Af3 8 �cl (or 8 Ag7). Evidently,
(16...�xg5 17 'i!txf3). On 15...Ae2, there follows Ljubojevic didn't feel like calculating the con­
16 'i!tg1! �e3 17 Ab6+ (the e2-square is occu­ sequences of 7... �xb2. However, as Villeneuve
pied). And if the bishop goes to g2, White plays has established, the bishop sacrifice is insuffi­
'i!te1! (analysis by Dvoretsky). cient: 8 'i!txd1 c4 (8 ... �xa3 9 �c2 �b4 10
�b2=) 9 Ag7+ c3 10 �e2! (10 Ae5? b5 11 Ad4
Ljubojevic-Karpov b4-+) 10...�xa3 (10...'i!tc2 11 'i!te3 b5 12 'i!td4!
Milan 1975 �d2 13 'i!tc5!) 11 Axc3 (11 �d3 b5 12 �c2!)
1l...b5 12 Ae5 b4 (!:::.. 13 ...'i!ta2-+) 13 Ad6!=.
7 Jl,g7? �c21
Only now, when the white king is cut off
5-25 from the queenside, does his position become
lost. Black's pawn advance will reach its goal -
but only with the black bishop on b3, which is
whereKarpov is sending it now.
w 8 Ae5 Ah5 9 Af6
9 Ac7 wouldn't help: 9...�xb2 10 Axb6
c4 11 Ad4+ (11 Ac5 c3 12 Ad4 �c2 13 Af6
Ae8 followed by 14 ... Ab5 and 15... 'i!tb3)
Of course, the position is drawn. All White 1l...�xa3 12 �d2 �b3 13 Af6 a3. Then Black
need do is to take the kingside pawns offthe light will place his bishop at b I, pawn at a2, transfer
squares, and his bishop can defend them. This his king to g6 and (with the white bishop at g7),
frees the king to counter Black's play on the trade the c4 and h6 pawns by means of ...c4-c3.
queenside, where he wants to create a passed 9 .•. Af7 to Ae5 Ah31 11 Ag7 b5 12
pawn. Jl,f8
The simplest solution to the problem is 1 Nothing would be changed with 12 Ac3 b4!
g5! f5 (l ...fg 2 hg !:::.. f4=) 2 f4 �d5 3 Ag7. An­ 13 Ag7 03 ab a3!; 13 Ae1 �xb2 14 ab cb 15
other reasonable line would be 1 h5!? g5 (l ...g h Axb4 a3 16 'i!td4 a2 17 Ac3+ �c2 18 Aa1 'i!tb1
2 g h �d5 3 Ag7 f 5 4 h 6 �c4 5 f 4 �b3 6 �d2= 19 Ac3 Af7! 20 �e5 Ag6-+) 13 ... c4.
Matanovic) 2 'i!te4 Ac2+ 3 �e3 f5 4 gf+ �xf5 12...c413 Jl,g7 b4!14 �d4
5 h6=. The main line ofKarpov's idea runs 14 ab
1 �e4?1 a4 2 h5? c3 15 Axc3 (15 be Ac4!) 15...a3 16 Ae5 a2-+.
White is doing all he possibly can to com­ Without the bishop at b3 in the final position,
plicate his life. Here again, 2 g5! f5+ 3 'i!te3 would White could save himself with 17 b3.
have secured an elementary draw. 14...c315be ha 16c4a2 17�c5�bl
2 gh 3gh f5+ 4 �e3�d5 5h6�c4
.•. 18 �b4 at tfJJ 19 Jl,xal �xal 20 c5 �b2
6f4�b3 21 c6 a3 22 c7 Ae6 23 �c5 a2 24 �d6
Jl,c8 White resigned.

104
 Opposite-Colored Bishops

Exercises

5-27

515
B?

The Bishop Restrains the Passed Pawn

Situations in which the bishop stops a 2...�e6, bringing the king closer to the impor­
passed pawn (and sometimes two - on the same tant f5-square.
diagonal) we call the second defensive system. 3a7Ae44g3?
The weaker side's king in these cases "maintains His opponent's incaccuracy remain unpun­

the zone" - that is, it defends its pawns, and ished. As John Nunn has correctly noted, 4 g4!!

limits the activity of the opposing king. hg 5 �g3 won. For example, 5 ... Af3 (with the
Attempts to break down the second defen­ king on e6, �f5 holds) 6 �f4 �e6 7 itd4 g6 8

sive system invariably involve breaking through h3 0 Ag2 (8...g5+ 9 �g3 +-) 9 �:g4! +- (but
not 9 hg Ah1 =)with a situation as soon arose in
to the passed pawn with the king (often after a
the game.
preliminary diversionary attack, and "widen­
Also sufficient for victory is 4 h4! g6 5 g4!
ing the beachhead" on the other wing).
hg 6 �g3 Af3 7 �f4 �e6 8 �g5 Ae4 9 �:g4
�d7 (the attempt to transfer the king to b7 is
Euwe-Yanofsky
hopelessly late) 10 �f4 Ag2 11 �g5 Ae4 12
Groningen 194 6
a8� itxa8 13 �:g6+- (shown by Burkhard
Treiber).
4... �e65�e3
5-28

5-29

B?

l...h5!
A typical move, ensuring the safety of the
kingside pawns. On 1 ...�e6?!, Black would have 5...Ag2 ?
had to reckon not only with 2 g4!?, but also with An instructive error: the white king should
2 �f2 �d7? 3 Af8 g6 4 Ah6!, when the h7-pawn not have been allowed near the pawns. The draw
becomes an attractive target for the white king. becomes unavoidable after 5... �f5! 6 Af8 g6 7
2 �f2 Ad3? �d4 Ag2 8 �c5 �e6! 9 �b6 �d7 10 b4 Aa8 11
A technique we have already seen more than b5 �c8! = (but not 11 ... Ag2? in view of 12 a8�!
once: the a-pawn is forced onto a square of the itxa8 13 �a7 Af3 14 �b8+-, with the unstop­
same color as its bishop. However, now was not pable threat ofb5-b6-b7).
:he time to attack the pawn. Necessary was 6�f4!g67g41

105
 Dvoretsky's Endgame Manual

The first step is to widen the kingside beach­ s. f49 Ag6Ae310 Ac2 h511 Af5c5
..

head. t2 Ag6h4!
7...hg 8 c;!!jlxg4 �h1 9 c;!!jlgS c;!!jlf710 �d4 Black only gets a draw out of12 g4? 13 hg
...

Ag2 11 h4 Aht t2 b4 Ag2 13bS Ahl hg (13...h4 14 .lle4) 14 fg, for example: 14...�xg4
15 '<fte2 �g3 16 <;f;>fl (but not 16 .ll e4? c2 17
.llxc2 �xg2) 16 ....llf2 17 .ll e4! c4 18 '<fte2! c2 19
'<ftd2=.
5-30 13Af5(13 Ae4 c40 ) 13...g4!14hg
No better is 14 fg f3 15 gf '<ftxh3 -+.
14 h315 gh �xf316 g5 �g317 g6
•..

Ad4 18 h4 f3 19 h5 Ag7 20 �et f2+

W? White resigned. After 21 �fl �f3, the king
marches unhindered to d2.
And now, let's examine a much more com­
plex ending, excellently played and annotated by
t4Af6!Ag2 Kaidanov.
On 14 ... .lle4, both 15 '<ftf4 !:::. '<fte5 and 15
b60 are strong. Kaidanov-Antoshin
lS h5!(the second, decisive step!) 15...gh RSFSR eh 1984
16 c;!!jlfS Black resigned.
If 16... '<fte8, then 17 '<fte6 !:::. '<ftd6-c7.
White's bishop restrains the h-pawn and simul­
5-32
taneously deprives the enemy king of the squares
e7 and d8 on the single diagonal d8-h4.

Makarychev-Averbakh
W?
Lvov 1973

5-31 What plan should White select? 1 M4? (hop­

ing to induce the reply 1.. .h5, giving his king
invasion squares on the kingside) would be a
gross blunder, in view of the pawn sacrifice l...g5!
B? 2 hg hg 3 .ll x g5 '<fte8. Black's king arrives at b7
(the " first defensive system"), and White is un­
able to create a second passed pawn on the
kingside.
Black's plan is the same as in the preceding
By the way, . g6-g5 is not yet a threat- White
. .

example: first, the king invades the kingside; then,

replies h4-h5, fixing the h6-pawn. ( With a light­
the beachhead is widened; and finally, the king
squared bishop, for the weaker side to have his
breaks through to the c-pawn.
pawns on dark squares renders them weak, and
t c;!!jle52 Ac2 �f43Abl Ah2 4 �f2
is generally a serious positional defect.) But with­
...

.ilgl +l 5�e2
out exchanging offthese pawns, it makes no sense
5 'it'xg1 'it'e3 6'i!tf1 'i!td2 would lose immediately.
to go into the first defensive position, because
s . �g36�fl Af2 1
the bishop will be unable to defend its kingside.
. .

In order to prepare ... t7-f5 Black must first

White will not be able to get to the a-pawn
,

take control of the et-square.

through the queenside: the enemy king will
7Ac2 f51 8 Abt
"maintain the zone." But by doing so, he will be
On 8 �xf5, the king gets through to his
diverted from the t7-square, and then White can
passed pawn: 8... '<ftf4 9 �c2 '<fte3-+ (it is impor­
play .ll g7, induce h6-h5, and return with his
...

tant that White cannot reply 10 �e1).

106
 Opposite-Colored Bishops

king to the kingside. Let's try it: 1 �d2 i.ta8 2 9...�g6(9.. .ef 10 gf e:. �f4,e4+-)lo
�c3 i.tb7 3 �b4 �e7 4 Ag7 h5 5 �c3. �h4Aa8 11 g4t fg 12 �xg4Ad513 �g3
Is there a way to prevent the king march via Having strengthened his kingside posi­
h2 to e5? Kaidanov suggests a counterattack by tion to the utmost, White brings the king over
Black's king: 5... �d6 6 �d2 �d5 7 �e1 �c4 8 to the queenside. Black must send his king to
�fl �d3 9 �g1 �e2=. However, he must also meet it - but then the g-pawn charges ahead.
consider 7 �e2 ! (instead of 7 �e1 ) 7... �c4 8 f3. 13...�f714 �f2 �e715 �e1 �d6
It would be safer to exploit the absence of 16�d2 Ac6(16 ... �c7 17 g6 �b7 18 g7 e5
White's king from the kingside by switching, at 19 de �xa7 20 f5+-) 17�c3 Aa8 18 �b4
precisely this moment, to the first defensive po­ Ad519 g6�e720 �c5 �f6(20 ...Aa8 21
sition: 5... �d7 6 �d2 Ad5 7 �e1 �c8 8 Af8 (8 f5 ef 2 2 d5+- ) 21 f51 Aa8 22 fe �xe623
i.te5 �b7 9 Ab8 changes nothing, while 8 �f l d5+ Black resigned.
�b 7 loses the pawn at a7) 8... �b7 9 Ac5 Ac4!,
and if lO f3, then 10 ...f4!!=.
1 �flt Aas 2 �g1 Ad5 3 �h2 Aas 4
�g3 1
Now let's examine 4 �h3 Ad5 5 g4? fg+ 6
�xg4 Aa8 7 �g3 Ad5 8 �g2 Aa8 9 �fl Jld5
10 �e2 Aa8 11 �d2 Ad5 12 �c3 Aa813 �b4.
Now, the defensive plan of marching the black
king down to e2 (13... �e7? 14 Ag7 h5 15 �c3 Exercises
�d6 16 �d2 �d5) doesn't work, because his
Both of the following exercises are rather
opponent will advance through the now open
difficult. In the first, you must calculate varia­
square g2. But there is another idea: Black can
tions accurately; in the second, you must find a
return to the first defensive position: 13 ...Ad5!
far from obvious plan of action.
14 �c5 �e7 15 Ag7 h5 15 �b4 �d6 17 �c3
�c7! 18 Ae5+ �b7 19 Ab8 �b6! 20 �d2 �b7
21 �e1 i.tc4! =, and White's king will not get to
the kingside.
4 Ad55 Ac7t
.•.
5-34

5/6
5-33 W?

5...�e7 5-35
Forced, because the temporizing 5...Aa8?
allows White's king to get to its passed pawn: 6
�f4! g5+ 7 �e5! gh (7 ... �e7 8 h5!+-) 8 517
�d6 +-. With the king already on e7, 6 �f4? W?
g5+! = no longer works for White; on the other
hand, the bishop sacrifice now becomes strong.
6Af4t g51 7Axg5+ t hg s hg �f79 f4t
But not9 �f4? �g6 10 f3 �f7 (or 10 ...�h5)
ll fe fe=.

107
 Dvoretsky's Endgame Manual

Chapter 6

Bishops of the Same Color

Minimal Material

Bishop and Pawn vs. Bishop

These endgames were first subj ected to 6-2

thorough analysis in the mid- 1 9th century by
the Italian player Centuri n i . Later, significant
a d d i t i o n s to the theory were made by GM
Averbakh . w

6-1 1 Ab7 Af5 2 Af3 Jtcs 3 Ae2o+­

All the squares on the c8-a6 diagonal, ex­
cept c8, are under the control of White pieces -
that's why we get a zugzwang. Now, if we were
to move the entire position down one rank, the
bishop would get another free square, and White
could no longer win.

White to move wins, by driving off the en­ T h e fol l ow i n g p o s i t i o n o f reciprocal

emy bishop from one diagonal, and then inter­ zugzwang has some practical significance.
fering along the other diagonal .
1 Ad7 Ad1 2 Ah3 Aa4 3 Ag2 1:::.. 4
Ac6+-
Can this plan be prevented? Yes, it can - 6-3
provided B lack's king can get to c5, preventing $
White 's bishop from interfering along the diago­
nal. Black to move draws:
1 . . . �d4! (but not l . . .�d5? 2 .ild7 Adl 3
.ilc6+ and 4 d7) 2 Ad7 J}.d1 3 Ah3 Aa4 4
Ag2 �c5! =
Thus, if the weaker side's king cannot get
in front of the pawn, then the basic defensive White to move draws . 1 Ad5 'it>c8 (or
principle becomes: king behind the king! 1.. .�a6) i s useless. On 1 Af5, there follows
The short diagonal: even with the "right" 1 Af3 2 Ae6 ( 6 3 Ad5+-) 2 Ab7! 3
••• •••

king pos ition, the draw is impossible, if one of �c5 Af3 (3 . . 'it>e7? 4 Ad5) 4 Ad5 Ae2
.

the diagonals along which the bishop will restrain ( 6�c8) 5 Ab7 �d7 =
the pawn proves too short. But what is B lack to do, if it is his move?
Any bishop retreat along the h 1 -a8 diagonal is
refuted by 2 Ad5; therefore, he must play
1 j},a6. By the way (here's a tragicomedy ! ), in
•••

thi s won position, Botvinnik accepted a draw

against Model in the 1 93 1 Leningrad Champi­
onship.
The path to victory i s uncomplicated : 2
�c6! Acs 3 Ac4! o Ag4 4 �b7! Af3+ 5
�a7 6 Aa6-b7+-.

1 08
 Bishops of the Same Color

Transposition to Positions White has nothing t o play for, other than to

with One Pawn pick up the b-pawn in exchange for his g6-pawn.
Unfortunately, this plan would not be enough to
C harushin - Rosenholz win. I present the main variation: 1 �c5 b4 2
er 1 986 �c4 .ll. e 1 3 .ll.c 5 �g7 4 .ll. x b4 .ll.g 3! (Averbakh 's
analysis shows that 4 . . . .1l.f2 also draws, but that
4 . . . .1l.h4? loses) 5 .ll.c 3+ �xg6 6 b4 �f7 7 b5
.ll.c 7 ! 8 �d5 �e7 9 �c6 �d8 1 0 �b7 �d7 =.
6-4 1 \t'e4
Capablanca is in no hurry to force matters
- he maneuvers, hoping for a mistake by his
opponent.
W? 1 ... b4
By no means forced ( l . . .Ae 1 2 �d3 .llb4
3 .ll.c 3 .ll e7 isn't bad); but, on the other hand, it
doesn 't spoil anything.
A typical s ituation : White can take the g4-
2Ae3Ac3 3 \t'd3Ae1 4 Ad2 Af2 5
pawn only at the cost of his a6-pawn. The ques­
\t'e4 (5 -'txb4 �g7=) 5 ....Q.c5?
tion is whether the enemy king can get back in
And here 's the mistake ! Now White cap­
time. tures the b4-pawn, w ith a tempo ahead of the
1 \t'f4! 0 other variations. First Black had to lure the king
Excellently played! White improves his own away from the queenside : 5 . . . �g7! 6 �f5 , and
king's position (now it no longer stands in the now he can defend the pawn (6 . . . Ac5 7 .ll f4 .ilf2
path of its pawn) while simultaneously using 8 .lle 5+ �g8 =).
zugzwang to force the enemy king further away 6 \t'd5! Ae7
from the kingside. The hasty 1 .ll. xg4? .ll. x a6 2 Still worse is 6 . . . .1l.f2 7 .ll. x b4 �g7 8 Ac3+
�f4 �c7 3 .ll.f3 �d6 4 g4 �e7 leads only to a �xg6 9 b4 �f7 1 0 .ll.d4 .llg3 1 1 b5 .ll c7 1 2 �c6
draw. .lla 5 13 .ll e 5 t::. Ac7+- .
1 ... \t'a7D ( l . . .�c7 2 a7 .ll.f3 3 .ll. x g4) 2 7 \t'c4 \t'g7 8 A x b4 AdS 9 Ac3+?
Axg4A x a6 3 Af3 \t'b6 White errs in return - although it's not at all
No better is 3 . . . .llc 8 4 .ll.e4 �b6 5 .ll.f5 . obvious. The win was 9 Ad2! - a variation we
4 g4 \t'c5 5 g 5 \t'd6 6 g6 \t'e6 shall examine later.
Nothing is altered by 6 . . . �e7 7 �g5 �f8 8 ...9 \t'x g 6 10 b4 \t'f5 1 1 \t'd5
�h6 .ll.c4 9 g7+ �g8 1 0 .ll.e4 !::. 1 1 .ll.h 7+ .
7 \t'g5 Ac4 8 g7
B lack resigned, in view of 8 . . . �f7 9 �h6
�f6 10 �h7 �g5 1 1 �h8 �h6 1 2 .ll.e 4, fol­ 6-6
lowed by .ll.h 7-g8 (the h7-g8 diagonal, where the
black bishop must move, is too short) .

Capablanca -Janowsky B?/Play

New York 1 9 1 6

In this position, Janowsky resigned. And

6-5 wrongly so - as Averbakh has shown - Black
could get a draw by employing the basic defen­
sive plan of "king behind king." Since White is
going to put his king on c6, B lack must hurry his
w king over to c4.
11 \t'f4!! 1 2Ad4 (12 .ll.e 5+ �e3 13 b5
••.

�d3 14 �c6 �c4=) 12 ... \t'f3! 13 b5 ( 1 3 .ll.c 5

1 09
 uvurt:LsKy s nnugamt: !V1anua1

�e2 14 �c6 r;ftd3 15 r;ftd7 .llg5 16 b5 �c4) Now we are looking at the position from the
13 . . . \!le2! 14 \!lc6 Wd3 15 Ab6 Ag5 16 next-to-last diagram, but with the bishop on d2
Ac7 Ae3 (instead ofc3). Here B lack's king is unable to get
After 17 Jld6 �c4, B lack has time to pre­ behind White 's.
vent the interference along the diagonal at c 5 . 1 l . . .�g4 1 2 b5 ®f3 13 �c6 �e4 1 4 ®b7 ! !
But the struggle is not over yet. �d3 1 5 Jle 1 ! �c4 1 6 �a6 �b3 1 7 Aa5 Ag5 1 8
17Wd5! b6+-.

Interference

We know that intereference is the primary

6- 7 instrument by which the stronger side secures (or
attempts to secure) the queening of its pawn. In
all the examples we have looked at thus far the
bishop has done this work. But sometimes (al­
B? though certainly not nearly as often), interfer­
ence is carried out with the aid of the pawns. For
instance, there is the following spectacular study.

T h e m o s t dangero u s c o n t i n u at i o n , as P. Heuicker, 1930

p o i nte d out by l s s l e r. I f B l ac k now p l ay s
1 7 . . . �c3?, then 1 8 Ad6 Jlb6 ( 1 8 . . . ®b3 1 9 �c5
®a4 20 ®c6) 19 ®c6. B lack hasn 't time to play
®c4 - White is ready to reply with either 20 �c7 6-8
or Ac5 , depending on where B l ack's bishop re-
treats.
B lack is saved by a tactic, which is very
useful to remember: it's a typical trick in bishop W?
endgames.
17 . . .Ad2!!
On 1 8 b6, the pin 1 8 . . . Aa5 saves him.
18Ad8Ae3! 1 Aa7! (1 h7? e4=) 1 ... Aa1 2 Wbl -'tc3 3
Now the threat of 1 9 b6 Jla5 20 b7 forces Wc2 Aa1 4 -'td4!! Axd4 ( 4 . . . ed 5 �d3+-) 5
Bl ack to retreat. That's fine - White 's bishop Wd3 -'l.b2 6 We4+-.
stands worse on d8 than it did on c7, and there is
no longer any danger in 19 �e7 ( .6. 20 Ac5)
19 . . . Jlb6! 20 �c6 Jla5 ! (White no longer has 2 1 We have already seen the tragicomedies that
.ilc7) 21 .Q.d6 �c4=. occurred in the games Botvinnik-Model and
White has j ust one final trap : C apablanca-Janowsky. I will add one more ex­
19 Ac7 Ad2! 20 Wc6 Ae3! 21 Wb7! ample.
( 2 1 Jld6 �c4=) 21 . . . \!}c4 22 Wa6Wb3!!
Savchenko - Krivonosov
Once again, the same technique of "king
USSR 1 989
behind king" : the black king heads for a4. He
would lose after 22 . . . Af2? 23 Jlb6 Jlh4 24 Jle3
.Q.d8 25 .Q.d2 .6. .Q.a5+-. And 22 . . . �b4? 23 .Q.b6
.Q.g5 24 .Q.a5 + and 25 b6+- is wrong too. 6-9
23 Ab6 Ag5 24 Af2 Ads 2 5 Ae1
Wa4 =
All that's left for us to see is what would
have happened, had Capablanca played more B
exactly on his 9th move.
9 �d2! ®xg6 1 0 b4 �f5 1 1 �d5

110
 Bishops of the Same Color

l .Q.e5?? 2 .Q.xe5 �d5 3 .Q.g7?? �c4!,

•••

and B lack won.

The same tactical idea of interference as in
6-13
the Heuacker study brought Black success here.
However, this occurred only as a result of his
opponent's gross blunder. After 3 <it>d3! <it>xe5
(3 . . . a2? 4 .llg 7+- ) 4 <it>c2, the king is in the
6/4
square of the a-pawn.
W?
B lack should have carried out his interfer­
ence in a more primitive form, by preparing
. . . .lle S. This could have been achieved either by
1 . . . <it>d5 2 <it>d3!? �e6! (but not 2 . . . a2? 3 c4 +) 3
�d4 a2 4 c4 .ll e 5, or by l. . .<it>d6 2 c4 .ll e 5
(2 . . . a2; 2 ... <it>e6) 3 c5 + <it>e6!-+ (3 .. .'itld5? is a
mistake, because of 4 c6= ). 6-14

Exercises

6/5
W?/Play
6-1 0

61 1
W?

6-11

6/2
B?

6-12

6/3
W?/Play

111
 Dvoretsky's Endgame Manual

The Bad Bishop

A vital principle of chess strategy (which is 11 hg �xe5 1 2 .ilg6! .ild5 1 3 .ilxe4 Ag8 14 �f4
certainly app licable in more places than the �d6 1 5 �f5 �e7 1 6 ®g6+- .
endgame) requires us not to place our pawns 4 b5 6 cb Axb5 6 Ac8 Ac6 7 b4 ab
•••

on the same eo/or squares as our own bishop. 8 ab Ab5 9 Ab7 g5!
In the first place, pawns that are fixed on On 9 . . . Ad3 , 10 Ae6 �f5 1 1 b5 (1 1 1.1.d7+)
the same col or squares as the bishop limit its mo­ 1 l. . .®g4 (ll. . .Axb5 1 2 1.1.xb5 ®g4 1 3 �f2 e3+
bility - this is why such a bishop is called "bad." 1 4 �g2+-) 12 b6 1.1.a6 1 3 �f2 e3+ 1 4 ®g2 is
In the second place, a bad bishop is unable decisive.
to attack the enemy pawns (which are usually 10 Axe4 gh 11 gh Aa4
placed on the opposite color squares), which 1 l. . .Ae8 loses also : 1 2 ilf3 �f5 1 3 Ae2!
dooms it to passive defense of its own pawns . (but not 13 �d4? �f4 and 1 4 . . . �g3) 13 . . . �e5
An d third, s inc e both pawns a n d bishop 14 l.td3 ! 0 Ad7 ( 1 4 . . . �d5 1 5 ®f4 �d4 1 6
control only one color of squares, there will be Ae2+-) 1 5 Ag6 ®d5 1 6 il.xh5 ®e4 1 7 l.te2+
"holes" in between those squares that the enemy �xb4 18 h5 l.tf5 1 9 ltd3 Ae6 20 h6 il.g8 2 1
pieces will occupy. �d4 .
12 Ag6 Adt l3 b5 \t'd5 14 \t'f4 \t'c5
Fixing Pawns 15 \t'g5 Ae2! ( 1 5 . . . ®xb5 16 il.xh5 il.e2 1 7
Ae8+ �e5 18 h 5 ®d6 1 9 �f6!+-)
Averbakh - Veresov
Moscow 1 947

6-1 6

6-15

W?

16Ae8!0
We know this technique from the ending
l h4! Charushin-Rosenholz (Diagram 6-4). Before tak­
The experienced player makes such moves ing the pawn, it is important to drive the black
- fixing the enemy pawns on the same color king back to b6 - as far as possible from the
squares as his bishop - without thinking. kingside. The hasty 16 il.xh5? il.xb5 17 il.g4
White has a great positional advantage. Af­ Ae8 1 8 ilf5 �d6 1 9 .llg 6 �e7! l eads only to a
ter the necessary preparations, he will create an draw.
outside passed pawn on the queenside, which will t6 \t'b6 17 A x h 5 A x b 5 18 Ag4
.••

Ae8 19 Af5 \t'c7 20 Ag6 \t'd8 21 \t'f6!

divert the enemy forces, allowing White to fall
Black resigned (analysis by Averbakh).
upon the kingside pawns.
t Ad7 2 Aft a5 3 Jlg2 Jlc6 (3 ... Af5
•••

4 Ah10) 4 Ah3!
The bishop aims for d7, where it will sup­
port the queenside pawn advance while at the
same time be ready to attack the pawn at g6. For
example: 4 . . . .ila8 5 .ild7 Ab7 6 b4 ab 7 ab Aa8
8 e5 be 9 be ®d5 10 .ile8 g5! (10 .. .'�xe5 1 1
.ilxg6 �d6 1 2 .ilxh5 �e5 1 3 Ag6 Ae6 14 g4+-)

1 12
 Bishops of the Same Calor

Zugzwang that much simpler. For example, 2 . . . g5 3 Ab2 Ad6

4 Ac1 Ae7 5 Ae3 Af6 C5 . . . Ad8 6 .lld 2 o ) 6 Ac5
With a bad bishop, the weaker side 's de­ Ad8 7 Aa3 Ab6 (7 . . .Af6 8 .llb 2 L:dtc3) 8 Ab2
fen s i v e hopes often are d e stroyed through Ac7 9 Ac3 0 . White 's bishop maneuvers here in
zugzwang. H ere 's the simplest example: roughly the same way as he did in the preceding
example.
Y. Averbakh, 1954 2 Ah41 .Q.e3
The c7-square turns out to correspond, not
just to the c3 -square, but also to g3 . 2 . . . Ac7?
would be bad : 3 Ag3! Ab8 4 Ae 1 Ac7 5 Ac3.
6-1 7 A n d on 2 ... Ad4 3 .lld8 decides.
3 Ag3 .Q.d4
After 3 . . .Af4 4 .lle 1 , Black must defend the
a5-pawn with his king, and allow the enemy king
w to enter. This bodes nothing good for B lack:
4 .. .'�b6 5 ®d5 'it'b5 6 Ac3 g5 7 AxeS .ll x e5 8
®xe5 ®b4 9 'it'd5 ®xb3 1 0 e5 a4 1 1 e6 a3 1 2
e 7 a 2 1 3 e8� a 1 � 1 4 �e3+ ®c2 1 5 �e2 + ,
The correspondence between the f3 - and
forcing the exchange o f queens.
f7-squares is obvious - to win, it is necessary
4 Ah21 o .Q.b2
only to give Black the move. If you like, you can
4 . . . Aa1 is even worse : 5 Ag1 Ab2 6 Af2
also find other pairs of corresponding squares
!:>. Ae1+- .
(for example, the fl- and b3-squares also corre­
5 .Q.gl .Q.a3
spond to f7), but there 's no real need.
On 5 . . . Ac 1 , there fol l ows 6 .ll f2 Ag5
1 Jle2 Jle8
(6 . . . Ad2 7 .llg3) 7 Ag3, and Black's bishop is
If l . . .Ag6, then 2 Ad3 Ah7 3 Afl ! .llg6
forced onto the f6-h8 diagonal - a fate which
(3 . . . Ag8 4 Ae2 Af7 5 Af3 o ) 4 Ag2 Af7 5
also befalls him in the game continuation.
Af3 o . 6 .Q.f2 .Q.e7
2 Jld3Ag6 Otherwise, we get the basic zugzwang po­
2 . . . Ad7 3 Ac2 Ae6 4 Ad1 Af7 5 Af3 o . sition: 6 . . .Ad6 7 Ae 1 Ac7 8 Ac3 0 , or 6 . . . Ab4
3 Jlc2 Ah7 4 Ab3! .Q.gs 5 Adt Af7 7 Ag3 Ad6 8 Ae 1 , etc.
5Af3 +-. 7 Ag31 Af6
By means of a series of accurate maneu­
Now, let's look at a considerably more com­
vers, Shabalov has achieved his aim - the bishop
plex endgame.
has been deflected onto a poor diagonal. On the
Shabalov - Varavin
other hand, there was no longer any choice:
Moscow 1986 7 ... .lld6 8 .Q.e 1 Ac7 9 .Q.c3 o +- .

6-18 6-19

w W?

1Ael .Q.b6 8 .Q.h2 0 .Q.g7 9 g51

On l. . .Ac7? 2 Ac3, Black is in zugzwang, White "breaks the rule," by moving a pawn
and must put another pawn on the same col or as onto a square the same col or as his own bishop­
his bishop, making his opponent's winning task in order to restrict the enemy bishop 's mobility

113
still further. There is no other way to reach his followed by . . . b7-b6. For example, 2 .llfl �d7 3
goal. �c3 �c5! (not allowing the enemy king to get to
9 ...Afs b4) 4 b4+ �d6. Here there can be no zugzwang,
9 . . . �h8 1 0 �g3 .llg7 1 1 .lle 1 is hopeless. since White 's bishop is unable to attack two en­
10 Axe5 Ae7 1 1 Af6 Ab4 1 2 Ac3 emy pawns simultaneously (as in the endings
Advancing the e-pawn does nothing for examined earlier).
White: 1 2 e5 .lld 2 1 3 e6 \t'd6 14 e7 \t'd7. So he 1 a4! g5
takes the a5-pawn in exchange for the g5-pawn. l . . . .lld 7! was more stubborn. On 2 �d4?
12 ...Ae7 13 A x a5 Axg5 14 b4 Af4 �xa4 3 .ll x d5 .ll c6 4 e4 g5 5 e5+ fe+ 6 fe+ �e7,
15 b5+ �d6 16 Ac31 g5 17 e5+ �c7 Black should get a draw. The right line would be
17 . . . �xe5 18 .ll x e5+ \t'xe5 1 9 b6! (but not 2 b3 �c5 (2 . . . b5 3 a5 �c5 4 b4+ �d6 5 �d4 is
19 \t'c5? \t'e6) 19 . . . \t'd6 20 \t'b5 g4 21 �a6+- . hopeless, in view of the weakness of the b7-pawn
18 Aa5+ �c8 19 �d5 g4 20 e6 g3 2 1 after the unavoidable e3-e4) 3 -'tf3! (3 b4+? �d6
�c6! Ag5 (22 e7 was threatened) 22 b 6 Black is premature). And now: 3 . . . g5 4 b4+ �d6 5
resigned. .lld 1 ! , with 6 �d4 to follow, leads to roughly the
same position as in the game. While 3 . . . h5!? gives
"Renegade" Pawns reasonable chances to survive.
2 �d4Af7 3 Af3Ae6 4 f5! Af7 5 b4
I n chess, there are no absolute laws. Even Ae8 6 b5!
so important and general ly useful an axiom as
the unprofitability of placing one's pawns on the
same calor squares as one's bishop must occa­
sionally be broken . H ere are the possible rea­ 6-21
sons for doing so :
- To restrict the mobility of the enemy bishop
using one's own pawns (as occurred in the pre­ B
ceding example);
- The need to undermine the enemy pawn
chain; and
- The attempt to create an impregnable for­
White's pawns have maximally restricted the
tress around a "bad bishop ."
enemy bishop. Now he brings his bishop around
The first and third points are illustrated by
to b3 , and plays e3-e4. When he thought up his
the following case:
plan, Wojtkiewicz had to calculate exactly the
pawn endgame that now arises by force.
Woj tkiewicz - Khalifman
6 ...Af7 7 Ad1 Ags s Ah3 Af7 9 e4
Rakvere 1993
Ags 10 Aa2 Af7 11 Axd5 Axd5 1 2 ed
�c7 13 �c3! (.6.1 4 �b4, 1 5 a5) 13 ... �d6
14 �c4 �e5
Also losing was 14 . . . �d7 15 �b4 �d6 16
6-20
a5 �xd5 ( 1 6 ... ba+ 1 7 �xa5 \t'xd5 1 8 �b6 �c4
19 �xb7 \t'xb5 20 �c7+-) 17 a6 ba 18 ba �c6
19 �a4 b5+ 20 �a5 .
15 a 51 ha 16 �c5 a4 17 d6 b6+ 18 �c6
W?
a3 19 d7 a2 20 dS� a1 � 21 �d6+ �e4 22
� xb6 �f3 23 �b7 �g2 24 �d3 �c1 25
b6 �c5 26 �b3 �h2 27 �f3 �d4 28 �c6!
� x h3 29 �c8 �b4 30 b7 �f8+ 31 �d7
The hackneyed 1 �d4? would have allowed �xg4 (31 . . .�£7+ 32 �d6 �f8+ 33 �e6) 32
Black to set up an impregnable fortress by l . . .b5!, �c8 Black resigned.

1 14
 Bishops of the Same Color

And now an example of the undermining 4 . . . a4 S ba ba 6 �d3 �dS 7 �c3 �e4 8 �b4 �f3
theme: 9 �xa4 �xg3 1 0 �bS �xf4 1 1 a4 gS 1 2 aS g4
Sveshnikov - Kasparov 1 3 a6 g3 14 a7 g2 1 S a8� g1i*, and the queen
USSR eh, Minsk 1979 endgame is completely hopeless for White.

Trauit:()medie�

6-22 Teichmann - Marshall

San Sebastian 1911

B?
6-23

First, let's evaluate what actually happened

in the game. B
l g60 2 <ifjle2 (the bishop can 't retreat,
•••

owing to 2 . . . �e 1 ) 2 Ac5 3 Axc5? (the pawn

•••

ending is lost) 3 <ifjlxc5 4 <it'd3 Cit'b4 5 <it'c2

•••

<it'a3 6 <ifjlbl a5 7 <it'al a4! (widening the Even though Black has an overwhelming
beachhead) 8 ba <ifjl x a4 9 <it'bl (9 �b2 b4) positional advantage, the endgame is not as
9 <it'a3 10 <it'al b4 11 <it'bl b3 White re­
•••
simple as it seems. Both sides made many er­
signed. rors; nor did grandmaster Averbakh avoid errors
White could have drawn by avoiding the in his commentaries.
exchange of bishops. After 3 �e 1 ! �e4 4 �aS, I l .i}.f7+?
.••

can 't see how Black can improve his position. An unfortunate move, allowing the king to
And if 3 . . . b4 (hoping for 4 .lld 2? �e4 S .lle 1 aS return to the defense of the kingside through the
6 .lld 2 .lld4 7 Ae 1 �e3, with zugzwang, or 7 d3-square. Now the position becomes drawn .
.llc 1 Ac3 8 Ae3 Ae 1 !), then simply 4 �f3!=. 2 <it'd3! <it'f4 3 Afl <it'g3 4 <it'e3 Ad5 5
But Black was the first to err here - the natu­ <it'e2 f5 6 <it'e3 .i}.e6
ral move 1 . . . g6? was a mistake. The pawn should The bishop sacrifice is insufficient: 6 . . .f4+
have been left on g7, in order to support the un­ 7 �e2 �b7 8 �e1 Axf3 9 gf �xf3 1 0 Ae2+
dermining with . . . t7-f6! The right way to obtain �g2 (10 . . .�g3 11 .llg4 �g2 12 �e2) 1 1 .llfl +
a zugzwang was by making a waiting move with �g3 1 2 �e2=. The only remaining try at mak­
the bishop. ing progress is . . . gS-g4, but this leads to the ex­
l . . .�aS! 2 �e2 (after 2 a3!? followed by change of too many pawns.
b3-b4, Black could also have tried for the win 7 <it'e2 g4
with the undermining . . . f7-f6 and . . . a6- a S )
2 . . . �e4 3 A c S f6! (undermining ! ) 4 e f gf. Black
continues by getting his bishop to c7 (or on S
Ad6 - to b6), his king to f5 , and playing . . . e6-eS
6-24
with a great and probably decisive advantage.
A reader found a second solution for this
position: l . . . .llc S!. If 2 Ae 1 b4 3 �e2 (3 Ad2
M2 4 �xb4 �xg3) 3 . . . �e4, Black wins using
w
one of the methods examined previously: either
by playing for zugzwang, or by undermining the
enemy pawn chain by t7-f6. Trading bishops also
loses : 2 �xcS �xcS 3 �c3 aS 4 a3 ( 4 �d3 �b4
S �c2 �a3 6 �b1 a4; 4 b4+ ab+ S �b3 f6) S hg

115
 Dvoretsky's Endgame Manual

Averbakh recommends 8 fg fg 9 �e3, which O n 2 .Q.fl , �f4 decides, for instance: 3 'it'd4
leads to an obvious draw after 9 . . . gh 10 gh �d7 f5! 0 4 'it'd5 �e3 5 �e6 �f2 6 .Q.c4 �xg2 , or 3
1 1 �e2 .\lb5+ 1 2 �e 1 �c6 1 3 �e2= . And if �d5 �g� 4 �e6 f5 5 'it'f6 �f2 6 .llc4 'it'xg2 7
9 . . . Ad7 then White replies either with Benko 's �xg5 �xh3 8 f4 ®g3- + .
recommendation of 1 0 �e2 .\lb5 + 1 1 'it'e1 Ac6 White has greater practical chances with 2
12 Ae2! (not 12 hg �xg2, when the h-pawn will Ad3!? Aa2+! 3 �c5 .
queen with check). Or with 1 0 hg! �xg4 1 1 .\lb5!
(pointed out by Cheron), giving up the g2-pawn
right away, but activating his bishop. For ex­
ample: 1 1 . . .Ae6 12 .\lc6 �c4 13 Ae4 Afl 1 4 6-25
.\ld5 .ilxg2 1 5 Ae6= (the attempted interference
leads to a drawn pawn ending), or 1 1 . . .�xg2 1 2
'it'f4 ! ( 1 2 .ll c 6 + ? � g 3 1::. h 3 - h 2 , .il h 3 - g 2 )
1 2 . . . .\le6 1 3 .llc 6+ �f2 (after 1 3 . . . 'it'h2 14 Ab7 B
h3 1 5 .\le4 �g1 16 �g3 h2 the interference on
g2 is impossible) 14 .ild5! Ad7 (14 . . . �xd5 1 5
'it'g4) 1 5 Ac6! .llh 3 1 6 .Q..d 5 Ag2 1 7 Ae6=.
Averbakh considers the text move the deci­ Averbakh contents himself with the single
sive error, but he's wrong. variation 3 . . . �f4 4 �d4 �g3-+ . But I think that
8 fg 9 'i!Je3?
...
3 . . . 'it'f4? is an error, owing to 4 �d6!
9 fg ! A x g 4 + 1 0 � e 1 ! w a s n e c e s s ary a) 4 . . . ®g3 5 �e7 �xg2 (5 . . .f5 6 �f6!) 6
(Averbakh only considers 10 �e3 Ad7- + ) , �xf6 �xf3 7 �xg5 �g3 8 �f5 ! (8 .llf5? Ac4,
leading to a curious position of reciprocal with . . . Afl xh3 to follow) 8 . . . .Q..d 5 (8 . . . 'it'xh3 9
zugzwang. White to move loses : 1 1 Ab5 �xg2 �f4=) 9 Afl ! (9 Ae4? Ac4 or 9 �e5? .Q.g2 1 0
12 Ac6+ �gl . But it's Black to move here, and �d4 Axh3 1 1 �e3 Ac8 1 2 .Q..fl Ab7 are both
after 10 . . . .\ld7 (10 . . . Ah5 1 1 Ab5 �xg2 1 2 Ad7, bad) 9 . . . .Q..c6 10 ®e5 Ad7 1 1 �e4 �f2 12 �f4
Ae6 1 3 .Q..b 5 Axh3 1 4 Ac6 (reaching a position
or 1 2 Ac6+ first) 1 1 .\la6 �xg2 (1 1 . . . .\lc6 1 2
from Cheron's line) 14 . . . Ac8 1 5 .Q..b 7! .Q..e 6 1 6
Ac8 Axg2 1 3 Ad7=) 1 2 Ab7+ 'it'g1 , White has
.Q.d5! , etc.
time to get his king to g3 : 13 �e2! h3 14 �f3 h2
b) 4 . . .f5 5 'it'e7 Ad5 6 Afl ! (6 �f6? is a
1 5 'it'g3=.
mistake, in view of6 . . . g4 7 fg fg 8 hg Axg2 9 g5
9 .1ld7?
h3 1 0 g6 h2 1 1 g7 .Q..d 5-+) 6 . . . g4 (6 . . . �e5 7
•••

B lack blunders in turn, allowing his oppo­

�d7 isn't dangerous either) 7 fg fg 8 hg ®xg4 9
nent to force the draw by the same means indi­
�f6 .Q..e 4 (9 . . . �g3 1 0 'it'g5 Ac6 1 1 ®h5=) 1 0
cated in the notes to move 8. The win was 9 . . . gf!
�e5! Aa8 1 1 �f6 .Q.b7 1 2 �g6 Ae4+ 1 3 �h6!=
1 0 gf .Q..d 7 0 1 1 'it'e2 (1 1 f4 Ag4! 0 1 2 �e4
(but not 1 3 'it'f6? �f4 ! , when White i s in
�f2-+) 1 1 . . .Ab5 + 1 2 �e 1 .llc 6 13 f4 Ae4! zugzwang).
( 1 3 . . . Ag2? 14 f5 h3 1 5 f6) 14 'it'e2 .Q..f5! 1 5 �e1 Black's king stands very well on e5, where
.ilg4 0 . it shoulders aside the enemy king. Before attack­
10 fg! .1lxg4 1 1 'i!]e4?? ing the g2-pawn, Black must first strengthen his
The loser is always the one who makes the position.
last mistake ! We already know that 1 1 Ab5 ! Simplest is 3 .. .f5!, for example: 4 <it'c6 g4!
would draw. But with the bishop o n fl, White i s 5 fg fg 6 hg Ad5+ 7 'it'c5 Axg2 8 g5 h3 9 g6
helpless. <it'f6!-+ or 4 .Q..fl �f4 5 �d6 (5 �d4 Ab1 ! 0 )
1 1 -'lcS 12 'i!]e3 -'ld7 0 White resigned.
...
5 . . ®g3 6 �e5 .Q..b 1 7 ®f6 ®f2 8 .\lc4 �xg2 9
On 13 �e4 (or 1 3 �e2), Black wins by �xg5 �xh3- + .
1 3 . . . .llc 6+ 14 �e3 .ll x g2; while if 1 3 �d2 �f2! And 3 . . . .Q.. e 6! 4 .\la6 f5 5 .Q.fl Ac8!? or
1 4 Ac4 �xg2 1 5 'it'e1 'it'g1 ! 1 6 �fl Ae6 0 1 7 5 . . . .Q..d 5 6 Ae2 .\lb7 7 M1 �f4 8 �d4 Ac8! 0 -+
.Q..b 5 h 3 1 8 Ac6 h 2 1 9 .Q..e4 Ah3 1::. .Q.g2 . are not bad either. However, the hasty 5 . . . g4?
Let's go back to the starting position of this would let slip the win: 6 fg fg 7 hg .\lxg4 8 .ila6!
endgame. Averbakh recommends l. .. Ab1 ! (on 8 �c4? Ac8! 9 �c3 'it'f4 10 'it'd2 �g3 1 1

1 16
 Bishops of the Same Calor

�e3 Ad7! 0 -+ or 1 1 �e1 Ag4! 0 -+ , we get

zugzwangs already fam il iar to us) 8 . . . 1te6
(8 . . . Af5 9 �c4) 9 Ab7! Af5 1 0 �c4 Ae4 1 1
6-2 7
Ac8=.

Euwe - Menchik
Hastings 1 930/3 1
W?

6-26
Comparing this position to the analogous
position after 2 . . . Axg4?, here Black has a passed
h-pawn. For this reason 7 �c4? Ae6! does not
W? win. On 7 �c3?! �e7 8 �b4, advancing the h­
pawn gives Black serious counterplay : 8 . . . h5! 9
�b5 h4 1 0 �b6 h3 1 1 �c7 ite6 1 2 .ll x b7 Ac4
t::.. Af1 -g2 .
White only keeps real winning chances by
White 's king wants to get to d4. Black will
7 e4! '1!e7 8 �e3! f6 9 f5 !, followed by �f4, and
prevent that with . . . �e5 ; after White responds
then either e4-e5 or �g4-h5.
with f2-f4+, he will try to counterattack with
7Af3?1
. . . �f5 and . . . �g4.
White restrains the passed pawn, but now
The accurate prophylactic move 1 ltb3!? is
Black's king is able to get to c7.
worth serious consideration. The idea is 1 . . . �e5
7 Cit'e7 8 Cit'c4 Cit'd8 (9 �b5 allows
2 f4+ �f5 3 Ad1 ! (covering g4 and intending 4
••.

9 ... �c7) 9 Cit'dS?I b6! 10 c6?!

�d3). For example, 3 . . . d4 4 Ac2+! �g4 5 ed
Euwe fails to sense the danger. He had to
�xg3 6 �e3 f5 7 Ad1 Ad5 8 Af3+ - ; or 3 . . . �g4
resign himself to a draw after 10 Ah5 .
4 �f2+ �f5 5 Af3! g5 (5 . . . �f6 6 �e2 �e7 7
1 0 Cit'c7 1 1 Cit'eS Ae6 1 2 fS Ab3
�d3 �d7 8 �d4 �c6 9 f5 Axf5 1 0 .ll x d5+ �c7
• • •

(12 . . . Ac4) 13 Cit'f6 bS 14 Cit'g7?

1 1 Axf7+-) 6 g4+ �f6 7 �e2! (7 f5 is possible,
The decisive mistake. 14 e4! would have
too) 7 . . . Ad7 8 �d3 Ac6 9 �d4 gf 10 ef �e6 1 1
given White the draw.
g5 �f5 1 2 g3+- .
14 b4-+ 1S Cit' x h7 Ac2 16 Cit'g7 b3
However, as Artur Yusupov has shown,
•••

17AdS b2 18Aa2 Cit'xc6 19 f6 Cit'd6 20 e4

Black successfully maintains his defensive posi­
A x e4 21 Cit' x f7 AdS + 22 JlxdS b1 � 23
tion with 3 . . . �f6! 4 �d2(d3) Af5 and 5 . . . Ae4.
Cit'g7 �g 1 + 24 Cit'f8 Cit'xdS White resigned.
The exchange ofbishops on e4 (after 5 Af3 ite4)
or f3 leads to a drawn pawn ending.
Exercises
1 Cit'd3 Cit'eS 2 g4
If 2 f4+ �f5 3 �e2 (3 �d4 �g4 4 .ll x d5
Axd5 5 �xd5 �xg3=) 5 . . . �g4 6 �f2 , then
6 . . . d4! (6 . . . h5 7 Ab3 g6 is also possible) 7 Axe6+ 6-28
fe 8 ed g6.
2 gSI
•••

Black loses after 2 . . . .1lxg4? 3 f4+ �e6 4 e4

�e7 5 Axd5 Ac8 6 �c4! 6/6
3 g3 A xg4 4 f4+ gf S gf+ (5 ef+ �e6 6 W?
�d4 Af3=) s .. Cit'f6 6 AxdS Ac8
.

1 17
 Dvoretsky's Endgame Manual

Barrier

Although there are occasional cases where Ad2 Ac3 6 Acl g4) 5 . . . Ac l 6 Ag1 g4 7 Af2 g3
a player can save himself with his pawns on the 8 Ag1 Ab2 9 '\t>d2 Ae5 1 0 \tle2 Ac3 0 -+ (Or
same color as his bishop, such a defensive 10 . . . Ac7 1 1 ®d2 .ilb6 1 2 \tle2 f4-+ ).
method is not to be recommended in the maj or­ 3 g4 4 g3
•••

ity of cases. The more secure defensive method If 4 Af2 g3 5 Ae l , Black "triangulates"
is to control the squares of one color with the with the bishop : 5 . . . Ad6! 6 ®d2 .ilc7! 7 'it>e2
bishop, and ofthe other color, with pawns. This .lle 5, and then wins the h4-pawn : 8 Ad2 (8 ®d2
places a barrier in the path of the enemy king, Ac3+) 8 . . . Af6 9 .lle 1 Axh4 10 \tld2 f4-+ . How­
making it difficult to invade our camp. ever, the text is no better.
If the opponent has a passed pawn, the king 4 Jtd6 5 Jl.f2Ac5 6Ag1 f4! 7 gf g 3
•••

must usually b lockade it. 8 f5 Ae7 9 <it>fl <it>f3! 10 e4 g 2 + 1 1 <it>e1

j},xh4+ 12 <it>d2 <it> x e4 13 <it>e2 Af6
I. Ivanov - Christiansen White resigned (14 \tlf2 Ad4+.)
Pasadena 1 983
In the following endgame, Dolmatov suc­
cessfully resolved much more complex problems.

6-29 Sveshnikov - Dolmatov

Yerevan zt 1982

W?
6-30

White 's position is difficult. The e3-pawn

greatly restricts the bishop 's mobility; and on the B?
other wing, the same role is played by the enemy
pawns (White 's bishop will not likely ever have
the opportunity to attack them from behind.)
Nevertheless, as Christiansen pointed out, 1 Aa6+!
•••

White has a comparatively simple way to draw ­ The weaker l . . . .ild7?! 2 Ag2 .ile8 3 '\t>d5
he must sacrifice a pawn, opening the diagonal .lld7 4 .llf3 0 .ile8 5 e6 would leave Black fac­
for his bishop and erecting an impassable bar­ ing the difficult problem of how to deal with
rier before the black king. threats on both wings (�d5-e5-f5 or b2-b3 , a3-
1 <it>d3! Ac3 2 e4+! fe+ 3 <it>e2 =
a4 and .llf3-e2-b5).
It's worth mentioning that the pawn endgame 2 <it>b3 Ab5!
after 2 Ad2? Axd2 3 �xd2 is lost: 3 . . . h4! (but not The king cannot be allowed to get to a4 -
3 . . . '\t>e4? 4 h4! f4 5 ef 'it>xf4 6 '\t>d3 �g3 7 \tlc4=) then White could secure the b5-square as well
4 ®d3 g6 0 (4 . . . �e5? 5 '\t>c4=; 4 . . . g5 5 '\t>d2 \tle4 by continuing b2-b3 and .ilc4 . Also risky is
6 \tle2 f4 7 ef gf!-+ ) 5 ®d2 '\t>e4 6 \tle2 f4 7 ef 2 . . . �c8 3 �c6 '\t>e6 4 \tla4 .ila6 5 Ab5 Ac8 6
'\t>xf4 8 'it>f2 'it>e4 9 \tle2 \tld4 10 ®d2 g5 0 -+ . Ac4+ \tlxe5 7 \tlb5 .
lvanov failed to find the pawn sacrifice, and 3 <it>c3
wound up in a hopeless position.
1 j},d2? <it>e4 2 J}.e1 g 5 3 h4
lf 3 Af2, then 3 . . . h4! 4 Ae 1 (4 Ag1 Ac7 5
Af2 Ab6 0 6 �d2 f4 7 'it>e2 Axe3 8 Ae 1 \tld4)
4 . . . Ab2 5 Af2 (5 g3 g4! 6 gh gh 7 Ag3 Ac1 ; 5

118
 Bishops of the Same Color

7Aa4Ag4 8Ac6 �e6

Of course not 8 . . . .lle 2? 9 .lld 5 and 10 .llc 4.
9 �b5 � xe5 10 � xb6 ,1ld1 1 1 h3
6-31 1 1 a4 .ilxa4 1 2 Axa4 �f4 13 .lld7 g4=.
ll . g4 12 hg .Q. xg4 13 a4 g5 14 a5
. .

Ae2 Drawn.

B? Setting up a barrier is an effective defen­

sive tool, but it too is not always sufficient. Some­
times the opponent can overcome the barrier by
offering an exchange of bishops. When doing
3 . . . �f8!!
so, it is necessary to calculate the pawn ending
A bri l l i ant defensive move di scovered
accurately.
through the method of exclusion. Let's follow
the grandmaster's logic .
Donner - Smyslov
The position after 3 . . . ltd7?! (or 3 . . . lte8?!)
Havana 1 964
4 �c4, with Ag2 and �d5 to follow, we have
already rated as unfavorable. In any event, it's
better not to choose such a course, if we don 't
have to.
6-32
3 . . . 1tfl ? loses to 4 b4! ab+ (otherwise, af­
ter the exchange of pawns on a5 , White 's king
obtains the important square c5) 5 �xb4, and
there is no defense against 6 Ac4.
B
3 . . . �e8? is bad: 4 b4! ab+ 5 'it'xb4 Ad7
6 e6 and 7 �b5 . 3 . . . �d7? fai l s for the same
reason .
Finally, on 3 ... 'it'd8?! there follows 4 Ac4
Ac6 5 Ag8! (threatening 6 'it'c4) 5 . . . Ab5 6 'it'd4, With the pawn on g2, Black could not have
and the king gets in via c4 or d5 . broken through the enemy defenses; but now it
But after 3 . . . 'it'f8!! 4 Ac4 Ac6, the g8-square is possible, although with considerable difficulty
is covered, and 5 .ll a 2 is not dangerous, in view - thanks to the weakness of the pawn at f3 .
of 5 . . . ®e7 6 �c4 'it'e6. 1 . . ..Q.h6+ 2 �c2 d3+ 3 �d1 �d4 4
4 b4 Af2+ �c3 5 .Q.b6 d2! 6 .Q.f2 �d3 7 Ab6
Before changing the contour of the game, .Q.f4 8 .Q.f2 .Q.e5 0 9 Ag1
White should have tried one more positional trap: lf 9 gh gh 1 0 Ag1 , then 1 0 . . . .1lc3 0 (pre­
4 1te6 �e7 5 .llc 8!? (cleverer than 5 Af5 Aa6! 6 mature would be 10 . . . .lld4 1 1 .llh 2, when Black
®d4-'l.b7!). The simplistic 5 . . . �d8? 6 Af5 ®e7 cannot play 1 1 . . .'it'e3 because of 1 2 .llg l + �xf3
would leave B lack in serious, perhaps insur­ 1 3 l;txd4=) 1 1 Ab6 .ll d4 1 2 .ll a 5 ( 1 2 .ll x d4
mountable, difficulties after 7 �d4 .lla6 (7 . . . .llc6 'it'xd4 13 �xd2 h4 14 �e2 f5 15 �d2 f4 0 )
8 Ae4 and 9 ®d5) 8 �e3! Ab7 9 �f2, with the 1 2 . . . ®e3 1 3 .ll x d2+ ®xf3 1 4 �e 1 ®g2 1 5 �e2
awful threat of �g3-g4. The threat of marching Ae5! b. .. .f5-f4-f3.
the king to g4 must be met by a timely transfer 9 h41
...

of the bishop to d5 : 5 . . . .\lfl ! 6 'it'd4 .llg2 7 �e3 Smyslov prepares the exchange ofbishops .
.)1d5! , and on 8 �f2 Black now has 8 . . .-'l.e6. The immediate 9 . . . .1ld4? leads only to a draw :
4 ab+ 5 �xb4Ad7 6Ab3 (White also
•.• 1 0 ltxd4 �xd4 1 1 �xd2 h4 1 2 g5!.
gets nothing from 6 e6 .lle 8 7 ®c4 �e7 8 �d4 10 .Q.f2Ac3 0 1 1 .Q.g1Ad41 1 2 .Q.xd4
'it'd6) 6 �e7
••• After 1 2 Ah2 Black sacrifices the bishop:
White can only seize the b5-square with his 1 2 . . . �e3! 1 3 .llg 1 + 'it'xf3 14-'l.xd4 �g2 1 5 'it'xd2
king by playing .Q.a4 first; and then Black's king �xh3 16 g5 ®g2 17 Ae5 h3 18 �e3 h2 19 Axh2
can attack the e5-pawn. �xh2-+ .

1 19
 Dvoretsky's Endgame Manual

12 ... �xd4 13 � xd2 �e5 14 �e3 g5 Black chose the desperate l ... dc? 2 -'l,xa8
White resigned, in view of 1 5 f4+ ( 1 5 '<fte2 cb, and after 3 -'l,e4 b2 4 h5 b4 5 �c4, he
'<ftf4 16 '<ftf2 f6 0 ) 15 . . . gf+ 1 6 'it'f3 f6 17 '<ftf2 '<fte4 resigned.
18 �e2 f3+ 19 �fl f2! (the standard triangula­ As Matanovic pointed out, Black could
tion maneuver, as seen in the game Fahrni - have saved the game by playing l . . .bc 2 be �c6
Alapin, doesn 't work here, since Black's king 3 .ilxd5 Ae8 4 c5+ '<ftc7. White 's king cannot
doesn 't have the f5-square available) 20 �xf2 get through the barrier.
(20 '<fte2 fl �+) 20 . . . '<ftf4o-+ .

Tf"asti£()medle�
Exercises
Matanovic - Uhlmann
Skopie 1 976

6-33 6-34

B? 6/7
B?

1 20
 Bishop VS. Knight

Chapter 7

Bishop versus Knight

With this configuration of material there is ing. For the practical player, what's important is
not, in my opinion, a single fundamental theo­ to become acquainted with the overall ideas, and
retical position that would be worth memoriz- with some concrete battle techniques.

Bishop and Pawn vs. Knight

V. Bron, 1955 conclusion : For a successful defense, it 's im­

portant to keep the knight far away from the
enemy king.

7-1 But even for the knight placed close to the

enemy king, zugzwang is not at all a sure thing.
Let's return once again to the diagrammed posi­
tion. Let's suppose that after 1 Ab3+ �c5 White,
w instead of the waiting move 2 Aa2!, chose 2 �e6?
4::l h 7! 3 �e7 (there 's nothing better) 3 . . . 4::lf8!

M. Mandelail, 1938
The outcome in all endgames of this sort
depends wholly on whether the stronger side
can place his opponent in zugzwang. In the
present case, this is possible. 7-2
1-'l.b3 + �c5 $
On 1 . . . �e5 2 .lle6 0 , the game ends at once.
2 -'l,a2 �c6 (2 . . . 4::lg4 3 �e6) 3 �e6 4)h7
(3 . . . �c5 4 -'tb1 or 4 Ad5) 4 -'ld5+ �c5 5 �e7 w
4)f6
5 . . . 4::lf8 6 Ae4 0 is no better.
6-'l.f3 4)g8+ 7 �e6 4)f6 8 Ae4! +­
The decisive zugzwang ! 4 .Q.c2 (White cannot allow the knight check
Let's put Black's king on e5 . Now the varia­ at g6) 4 ... �c6 (4 . . . �d5 is possible too) 5 .Q.a4+
tions are different, but the evaluation of the po­ With the bishop on the bl-h7 diagonal, the
sition doesn 't change, as well as the goal of king will shuttle between c6 and d5, avoiding
White's maneuvers - zugzwang. the mined squares c5 and e5 . For example: 5 Ab1
1 -'l.b3 �f5 2 -'l,f7 �g5 (2 . . . �e5 3 �d5! 6 Ad3 �c6! (6 . . . �c5? 7 Ae4 0 , 6 . . . �e5?
Ae6 0 ) 3 Ae6 �g6 4 �f8! 4)h7+ (4 . . . �h6 7 Ae4 0 ) 7 Ae4+ �c5 0 .
5 �f7 �g5 6 -'th3 0 ) 5 �e81 4)f6+ 6 �e7 0 5 ... �c5 6 .Q.e8 �d5 7 .Q.f7+ �c6 8
(in order to give his opponent the move, White .Q.h5 �c5! =
has triangulated with his king) 6 ... �g7 7Af7 8 . . . �d5? is a mistake, in view of 9 .llf3+
4)g41 8 Ad5 (but not 8 d7? 4::l e 5 9 d8� 4::l c 6+) �e5(c5) 10 Ae4 o +- . But now we have a po­
8 ... 4)e5 (8 . . . 4::lf6 9 Ae4! 4::lg8+ 10 �e6 4::l f6 1 1 sition of reciprocal zugzwang, with White to
Af5 0 ) 9 .Q.e4 �g8 10 �e6 4)f7 1 1 d7 �f8 move; and he cannot give the move back to his
12 Ad5 4)d8+ 13 �d6 0 �g7 14 �e7 +-. opponent.

Now, in the diagrammed position, let's move It is not uncommon in such situations for
White 's king to c7. It's not hard to see that Black Black to have a passed pawn, too. The stronger
can draw this - and not just with his knight on side's strategy remains unchanged: White must
f6, but also on f8 or e5 . Which brings us to the still play for zugzwang. The defender, however,

121
 Dvoretsky's Endgame Manual

now has a new resource: deflection. Sometimes, 1 hS!

the pawn distracts the bishop from control ling Exploiting the fact that the pawn is tempo­
an important square, which the knight then im­ rarily poisoned ( 1 . . .Axh5? 2 �d3= ), to advance
mediately occupies. Or the reverse can happen: it further. On 1 �e2? c;!td4 the position is lost.
sometimes the knight is sacrificed to allow the 1 .Q.h7 2 h6 �cS 3 �e2 �d4 4 �d1
•••

pawn to queen. �c3

Let's examine the other attempts to play for
Lisitsyn - Zagorovsky zugzwang:
Leningrad 1 953 4 . . . Ad3 5 �e 1 �e3 works well if White
plays 6 �d1? �f2 0 . But White saves himself if
he sacrifices a pawn to deflect the bishop: 6 h7!
.ll x h7 7 .tlc4+.
7-3 4 ... �d3 5 �e l �c3 6 �e2! (but not 6 c;!td1?
Ad3 7 c;!tel �c2 0 ). Now 6 ... Ag6 is met by 7
�e l ! , leading back to the game line (7 �e3?
.ll.d 3 0 or 7 �d1 ? Ad3 8 �el �c2 0 would be
W? a mistake). And if 6 . . . �c2 (counting on 7 c;!te 1?
Ad3 0 or 7 �e3? �d1 0 ), then White saves him­
self with the knight sacri fi c e indicated by
Konstantinopolsky: 7 .tlc4!! .lld 3+ 8 c;!te3 Axc4
9 h7 b1f! 10 h8fff = .
1 .ll.e8? would be a mistake in view of l . . .h5!
(deflection) 2 �xh5 .tld7+ 3 c;!te6 .tlb6. The S �e1 �c2 6 �e2Ad3+ 7 �e1!
knight stands far away from the king, and the 7 �e3? c;!tc3 0 is a mistake.
position would be drawn. 7 �c1 8 �b3 + �b1 9 �d1!
••.

1 .ilfS! hS 2 �e6 h4 3 �f6 0 �c6 The final touch. 9 .tld2+? �c2 0 or 9 c;!td2?
3 . . . h3 is no better: 4 Axh3 c;!te4 5 �e6 �d4 �a2 10 .tie l + �a3 both lose.
6 .ll.f5 0 . 9 .1}.c2 + 10 �e2 Ah7 11 �d1 .Q.c2 +
•••

4 d7 �d8 s j}_e6 ( b. .ll.d 5, �e7) s �e4...

1 2 � e 2Ag 6 13 � d 1 AhS+ 14 � d 2 �a2
6 -'lh3 15 h7 b1� 16 �cl + �a3 17 h8� Draw
The most accurate move, threatening 7
�g2+ and 8 c;!te7. But the immediate 6 �e7 .tlb7 Exercises
7 .ll.c4 h3 8 .ll.a 6+- was also possible.
6 ... �{3 7 �e7 �b7 (7 . . . .tlc6+ 8 �d6
In the following exercises, you must answer
.tld8 9 �c7 .tl£7 10 Ae6+- ) 8 -'tf1 � g 3 9
the question, "What should be the result of this
.1la6 �cS 10 d8� Black resigned.
game?''
Now, here 's a more complex example.

N azarevsky - Simonenko
Kiev 1 939 7-5

7-4 7/ 1
$ W?

1 22
 Bishop vs. Knight

7-6

7/2
W?

Knight and Pawn vs. Bishop

The bishop is a strong piece, sometimes These examples show us the two basic tech­
capable of preventing a pawn from queening niques for promoting the pawn: driving the
even without the king 's help. bishop off the diagonal, and interference.

If the bishop can 't handle the job on its own

(which is what happens most often), then the
7- 7 outcome depends upon the position of the de­
fending king : can it prevent the bishop from be­
ing interfered with or driven off?

B Y. Averbakh, 1958

1 . . . -'l,eS! (White threatened interference 7-9

with 2 .f:lc6) 2 4)d7 �g2 3 �d8Ag6 4 �e7
Af5 5 4Jc5 ( !::. .f:le6) 5 . . . -'l,cSI
=

Black was saved, first of all, because the

pawn had not yet reached the 7th rank, and sec­ w
ond, because the bishop 's diagonal was suffi­
ciently long: 5 squares. Knight and king are only
capable of interdicting two squares apiece, which
leaves the fifth square free. With the king at a I or b I , White would win
If we move the position one file to the left, by �d5-c4-b5-a6-b7. But here (or with the king
the diagonal grows shorter, and Black loses. at c l , also), Black's king is in time to help the
bishop: 1 �d5 �c3! 2 �c5 �d31 3 �b5
�e4 4 �a6 �d5 5 �b7 �d6 =

Note Black's accurate first move : l . . .�a3?

7-8 would be refuted by 2 �c4! �a4 3 �c5 o +­
(this is how White wins if the king is on a2 in the
starting position) . And if l . . .�b3? (hoping for 2
�c5? �a4, when White 's the one in zugzwang),
B then 2 .f:ld4+ �b4 3 .f:le6 Aa5 4 �c6, with the
unstoppable threat of interference by 5 .f:lc7.
The other plan, 1 �e7 �c3(b3) 2 .f:ld8 �c4
3 .f:le6 .Q.a5 4 �d6 ..llb4+! doesn 't win for White
1 . -'l,dS 2 4) f4 �f2 3 4)e6 -'l.a5 4
. . either. It's important that White doesn 't have the
�a6+- . 5 .f:lc5 interference - the king covers that square.

1 23
 Dvoretsky's Endgame Manual

But with the king at c 1 in the starting posi­ However, Dorfman played too straightforwardly,
tion, this plan leads to the queening of the pawn: and was unable to gain the point.
1 �e7! �b2 ( l . . .�d2 2 4Jd4! �e3 3 4Je6 .ilg3 4 •• l .Jild3 2 AalD (Black threatened the
�e8 .ilh4 5 4Jf8 �e4 6 4Jg6 !:;. 7 4Je7) 2 4Jd4! interference 2 . . . &Llb2) 2 �b2 3 <itlel
•••

.fta5 3 4Je6 �c3 4 �d6 Jlb4+ 5 4Jc5! l;ta5 6 3 'it'e3 must be answered by 3 &Lla4! 4 <it1e2
. . .

4Jb7 .ilb6 �c6+- , or 3 . . . .ilb4+ 4 �f6! .ilc3+ 5 �cl ! (see below), or 5 'it'd4 �bl 5 'it'd3 'Llc5+ 6
�f5 Jla5 6 �e4 �b3 7 �d5 !:;. �c6, 4Jc7+-. �c3 �xa l -+ . Whereas, in the game Sakaev­
Sunye Neto (Sao Paulo 1 99 1 ), after 3 . . . �bl ? 4
Tr-aeiwmedies 'it'd2! the win was gone.
3 <itlbl 4 <itld2 <itl x a l 5 <itlcl ! �c4 6
•••

Stein - Dorfman
<itlc2 Draw. We know the concluding position
from the chapter "Knight vs. Pawns" (diagram
USSR 1 970
2-2).
The road to victory was noted as far back
as the 1 9th century by Horwitz. Black should
have played 3 . . . &Lla4! (instead of3 . . . �bl ?) 4 <it1e2
7-1 0
�c l . Possible variations are :
$
5 �d3 �bl 6 �d2 'Llb2 0 7 'it'c3 �xal 8
�c2 'Lld3 0 -+ ;
5 � e 3 'it'b l 6 'it'd3 (6 � d 2 'Llb2 0 )
B 6 . . . &Llc5+! (of course not 6 . �xa l ? 7 �c2=) 7
. .

<it1 c 3 (7 � d 2 'Ll b 3 + ) 7 . . 'it' x a l 8 � c 2

.

'Llb3(d3) 0 -+ ;
5 'it'el &Llc5! 6 'it'e2 (6 �g7 'Lld3+ and
The bishop has a hard time with a rook's 7 . . . &Llb2) 6 . . . �bl 7 ®dl (7 11.g7 &Lla4) 7 'Lld3
. . .

pawn, since it has only one diagonal to work with. 8 �d2 'Llb2 0 -+ .

Exercises

7-11 7-12

7/3 7/4
W? W?

White to move - what result?

1 24
 Bishop vs. Knight

The Bishop is Superior to the Knight

Cutting the Knight Off Trastif:()medie§

If the knight is on the edge of the board, the Bykova - Volpert

bishop can deprive it of moves . USSR 1 95 1

7-13 7-15

w B

1Ae5! C3;e7 2 ®c5 ®d7 3 d6 (but not 3 1 4)e8??

...

<it>b6?? f:lf6!=) 3 C3;e6 4 C3;c6 C3; xe5 5 d7 +-.

...
A mistake that's hard to explain. Almost any
other retreat by the knight would have led to an
Sometimes it is not necessary to "arrest" uncomplicated draw. Now Black loses.
the knight - it's enough to cut it off from the main 2 ®g5 4)g7 3 C3;h6
theater of conflict (for example, from the passed Black resigned, in view of 3 . . . <£le8 4 gS 0 .
pawn), as in the following example.
Exercises
Goldberg - Tolush
USSR chsf, Moscow 1 949

7-1 6

7-14

7/5
W?
W?

1 h4! gh 2 gh 4)e5 3 Af5!

The bishop deprives the knight of the im­
portant squares g4 and d7, which it would other­ 7-1 7
wise use for the fight against the h-pawn. It is
true that the knight can immediately remove this
pawn - but then it comes "under arrest."
3 4)f3+ 4 C3;f2 � xh4 5 Ae4! C3;c7 6
•••
7/6
C3;g3 �g6 7 j},xg6 fg 8 f7 Black resigned
W?/Play

1 25
 Dvoretsky's Endgame Manual

Fixing the Pawns For now, White 's king cannot penetrate the
kingside: 8 �h3 .£)a5 9 .ild6 .£)b7 1 0 �e7+ �h5 .
It is useful to ftx the enemy pawns on 8 'i!7fS 9 Ag7 hS
•••

squares where they may be attacked by the 9 . . .�g5 is met by 1 0 �e2, when the h-pawn
bishop. In this case either the king or the knight must be advanced anyway. After h6-h5, White
w i l l be tied down to their defense. changes her plan, and decides the outcome on
the kingside.
Chibu rdanidze - Muresan 10 'i!7 g 21 4)cS 1 1 Af8 4)b7 1 2 'i!7h3
Lucerne oi I 982
�g.S 13 Jle7+ �f.s 14 �h4
Black resigned, since her king cannot si­
multaneously defend the pawns at h5 and f4.
7- 1 8
There can be no help from her knight, either - as
before, it's tied to the queenside; meanwhile,
throughout this endgame, White 's bishop re­
mained very active.
W?
Miles - Dzhindzhikhashvili
Tilburg 1 978

In order to make progress, White must bring

her king to the queenside - but this will be met
7-1 9
by the black king coming to d5 . For example: 1
Wf2? g6! 2 �e2 'it'e6 3 'it'd3 'it'd5 = . White also
gets noth ing from 2 b6 'it'e6 3 Af8 h 5 4 gh gh 5
'it'g2 'it'd7! 6 'it'h3 <£lc4 = .
w
One ofthe most important methods ofcon­
verting one 's advantage in endgames (and not
just in endgames) is "the principle oftwo weak­
nesses. " Sometimes it is impossible to win by
Whereas in the preceding example Black's
working only on one part ofthe board. In such
king was forced to defend its pawns, here this
cases, the attacking side strives to create a sec­
role is played by the knight. In order to let his
ond weakness in the enemy camp, or to exploit
king break into the enemy camp, White uses the
one which already exists. By attacking this sec­
standard techniques of widening the beachead
ond weakness, and then if necessary returning
and zugzwang.
the attack to the first weakness, the attacker suc­
1 g41 hg 2 fg l 'i!7g 6 3 'i!7g3 'i!7g S
ceeds in breaking down and eventually over­
On 3 . . .f5 4 gf+ ef 5 �f4! decides.
coming the enemy 's resistance.
1 h4! 4 'i!7f3 'i!7h6
An excellent positional move, stemming Nor does 4 . . . e5 help, in view of 5 �g3 �g6
from the "principle of two weaknesses." The vul­ 6 �h4 �h6 7 .i.td8 �g6 8 g5! fg+ (8 .. .f5 9 ef+
nerab il ity of the h6-pawn prevents Black's king �xf5 10 �h5 e4 1 1 g6+-) 9 .ilxg5 .£)b8 10 �d8
from heading towards the center; but how, then, .£)d7 1 1 �g4 0 +- .
is she to meet the advance of the enemy king to S 'i!7f4 <i!7g6 6 eSI fe+ (6 . . .f5 7 gf+ ef 8
the queen 's wing? e6) 7 AxeS 'i!7f7 8 Ac7 'i!7f6 9 g S+ 'i!7f7 10
l. . g6 2 h5! gh 3 gh
.
'i!7g4 'i!7g 6 1 1 .Q.d6
White's position is now won. Black resigned. He is in zugzwang, and will
3 ... 'i!7f6 4 b6 4)b7 5 Af8 <i!7gS 6 Ag7 find himself in zugzwang over and over again,
'i!7 x hS 7 AxeS since his knight is tied to the defense of the b6-
The h5-pawn is gone, but now the king must pawn, and cannot stir. For example : 1 1 . . .e5 1 2
defend another vulnerable pawn - the one at f4. Ac7 0 e 4 1 3 �f4 ( 1 3 .i.td8) 1 3 . . . e 3 14 �xe3
7 'i!7g S 8 'i!7f2
••• �xg5 15 �e4+ - , or 1 1 . . .�f7 12 �h5 �g7 1 3

1 26
 Bishop vs. Knight

Ac7 0 �h7 ( 1 3 . . . e5 14 �g4+-) 14 g6+ �g7 ( G o l d b e rg - To l u s h , fo r i n s tanc e , or

1 5 �g5 o +- . Chiburdanidze - Muresan). Let's analyze some
more examples of this theme.
Traui�umedies
Spassky - Fischer
Smyslov - Gurgenidze Santa Monica 1 966
USSR ch, Tbilisi 1 966

7-21
7-20

B?
W?

White would certainly love to play g2-g4

White wins, using exactly the same move (for instance, in reply to 1 . . .<iJd6), tying one of
(and the same technique) as in the Chiburdanidze­ the enemy pieces to the kingside. Then the king
Muresan game: 1 h4! It is vital to fix the enemy would move over to the queenside, and attack
pawn on the vulnerable h5-square, in order to tie Black's pawns.
one of Black's pieces to its defense, or in some The most stubborn line was Gligoric's sug­
lines to create a dangerous passed h-pawn. gestion l . . .<iJh6! (and if 2 'it>f4, then 2 . . . <iJf7! 3
In the game, White erred with 1 Cjf}d5? After g4 g5+ ). Averbakh extends the line as follows: 2
1 h4!, the position became drawn. If White
•••
�d3 <iJf5 3 �c4 <iJxh4 4 'it>xc5 'it>e5 5 Ab7 'it>f4
sends his king after the a7-pawn, Black squeezes 6 �b5 'it>g3 7 �a6 <iJxg2 8 �xa7+- (the knight
it into the corner with .. .'!lc7. And on g2-g3, is, as usual, helpless against a rook's pawn).
Black exchanges pawns and easily blockades the But instead of the desperate king march to
passed g-pawn which results. Besides, he only the g2-pawn, Zviagintsev suggested the more re­
needs to give up his knight for it, and then bring strained plan of 5 . . . <iJf5, which offers B lack re­
his king back to b8 (the elementary fortress al­ alistic saving chances, in view of the small
ready known to us) to secure the draw. amount of remaining material. On 6 'it>b5 there
The continuation was : 2Ae2 4)f8 3 Cjf}e4 follows 6 . . . �d6 7 �a6 'it>c5 8 �xa7 �b4= (af­
Cjf}g 5 4 Cjf}d5 Cjf}f6 5Ag4 4) g 6, and the game ter the king gets to a3 , the knight will be given
ended in a draw. up for the g-pawn). Or 6 a4 <iJe3 7 �b5 �d6 8
a5 (8 �a6 �c5 � 9 . . . �b4) 8 . . . g5 9 Ae4 g4 1 0
The Passed Pawn 'it>a6 g 3 1 1 'it>xa7 �c7 1 2 �a6 <iJc4 1 3 ltf3 <iJe3
14 'it>b5 �b8 1 5 'it>c5 ( 1 5 'it>b6 <iJc4+) 15 . . . �a7
Th e presence of passed pawns on the 16 �d4 <iJxg2=.
board, as a rule,favors the side with the bishop. Fischer 's choice makes things considerably
The bishop is a wide-ranging piece, able both to easier for White, since it gives him a passed pawn
support its own pawns, while simultaneously without even having to exchange pawns for it.
dealing with the enemy 's, whereas the knight 1 g 5? 2 h5 4)h6 3 Cjf}d3 Cjf}e5 4 .1l,a8
•••

generally succeeds in acting only upon a narrow Cjf}d6 5 Cjf}c4 g4 6 a4

segment of the board . If it succeeds, let's say, in Black's king can only defend one of the two
blockading the passed pawn on one wing, it can­ queenside pawns. Seeing that the a7-pawn is
not successfully involve itself in the fray on the doomed, Spassky does not hurry to attack it, pre­
opposite wing. ferring to strengthen his position maximally first.
A few of the endings we have examined 6 4) g8 7 a5 4)h6 8 .1l,e4 g3 9 Cjf}b5 4) g8
•••

have already illustrated the difficulties faced by 10Ab1 4)h6 1 1 Cjf}a6 Cjf}c6 1 2Aa2 Black re­
the knight when battling against a passed pawn signed.

1 27
 Dvoretsky's Endgame Manual

The following sharp endgame features an In Chapter 1 2, which is devoted to the theory
intere sting, though not w h o l l y error- free, of queen endgames, you will read that in such
struggle. situations the only hope for salvation lies in the
black king getting as close as possible to the
Perelstein - Vepkhvishvili corner square al. Black should therefore play
Pushkin Hills 1 977 8 . . . 'it>c5! 9 i1i'c8+ 'it>b4 10 i1i'xc2 i1i'h l + 1 1 ®g7
i1i'xf3. The computer assures us that the result­
ing position is drawn; however, to demonstrate
this evaluation right at the board is quite diffi­
7-22
cult - as a rule, the defending side errs some­
where along the way, and loses.
4 h5? gh 5 g5?
White still draws after 5 gh! .f:lxa1 6 h6.
w
s �e31
...

Now it's Black who wins. On 6 g6 .f:lf5 the

knight will sacrifice itself for the g-pawn, and
Who stands better? The black a-pawn could the b i shop cannot stop both passed pawns
become very dangerous, while White will soon ("pants").
create a kingside passed pawn. In such sharp po­ 6 <if}g6 h4 7 <if}f6 <if}d6! 8 g 6 �d5+ 9
sitions, the bishop is usually stronger than the <if}f7 �e7 10 g7 �f5?1
knight, which is why Perelstein didn't go in for 1 0 . . . h3 would have reached the goal a lot
the drawing line 1 h5 gh 2 'it>xh5 .f:le3 3 g4 a3 4 more simply.
�h6 .f:lc2 5 �xh7! .f:lxa1 6 g5 a2 7 g6=. 1 1 g8�1 h3 12 �f6 h2
1 <if} h 6 � e 3 2 g4 a3?1 There was also a more elegant solution,
The accurate 2 . . . .f:lg2! 3 h5 gh would have based on the idea of a deflecting knight sacri­
led to a draw. Black hopes for more, and does fi c e : 1 2 . . . � c 6 ! 1 3 .f:lg4 ( 1 3 .f:l e 4 .f:ld6 + ! )
indeed achieve it - but only as a result of errors 1 3 . . . .f:lh6+! 1 4 .f:lxh6 h2- + .
on the part of his opponent. 13 �e4+ <it'd5 1 4 � f2 �d4
3 <if}xh7 �c2 As a result of the inaccuracies committed,
Already pointless is 3 . . . .f:lg2 4 h5 gh 5 gh. the knight must repeat his earlier task of attack­
ing the bishop, and then returning to battle with
the enemy passed pawns. He turns out to be just
in time.
7-23 1 5 f4 �c2 16 f5 (there's nothing else)
16 � xa1 17 <if}e7 �b3 18 f6 �d4
•••

White resigned, in view of 1 9 f7 .f:le6

(analysis by Dvoretsky).
W?
Exercises

The bishop can find no square on the long

diagonal: 4 .llf6? loses to the interference move 7-24
4 . . . .f:ld4, and 4 Ac3? is met by 4 . . . 'it>c4. All that's
left to hope for is his pawns.
The strongest move here was 4 g5!! After
4 . . . .f:lxa1 5 h5 a2 6 hg .f:lc2 7 g7 a1 i1i' 8 g8i1i'+ 7/7
Black loses his knight. For example : 8 . . . �d6 W?
(8 . . . �d4? 9 i1i'g7+) 9 i1i'g6+ �e7 1 0 i1i'xc2 'l*h1 +
1 1 �g7 i1i'xf3 1 2 i1i'c7+ 'it>e6 1 3 i1i'b6+ �d7 1 4
i1i'f6, with a winning queen endgame. Can White save himself?

1 28
 Bishop vs. Knight

An Open Position, A More Active King ( ti. l 4 . . . �g4) 14 f3 gS l S hg �xgS!, and the king
reaches g3 .
The classic example of the exploitation of 12 h5 13 �f6+ �f5 14 �d7?
•••

this type of advantage is the following endgame. Once again Stoltz fails to show defensive
grit. As MUller and Lamprecht indicate, Black's
Stoltz - Kashdan task would have been considerably more diffi­
The Hague ol 1 928 cult after 1 4 <£\h7! �g4 (14 . . .Jlfl lS f3, and if
1S . . . Axg2, then 1 6 �e2) lS <£\f8 �xh4 16 <£\xg6+
�gS 17 <tieS �fS 18 <£\f3 (the pawn endgame
after 18 <£\d3 Jlxd3 19 �xd3 �g4 20 �e2 h4!
7-25 21 b4 a6 22 �fl �fS 23 �e l �e S ! is lost)
1 8 . . . Ab7 1 9 �e2.
14 -'l,c8!
•••

Excellent technique. On l S 4JcS �g4 de-

B cides; however, the text is no improvement.
15 �f8 g5! 16 g3
Forced: after 16 hg �xgS the knight is lost.
The position seems about equal, but it is 16 gh 17 gh �g4 18 �g6 Af5 19
•••

�e7 Jl,e6 20 b4 � x h4 21 �d3 � g4 22

not: Black has a significant advantage, in fact.
First, because his king succeeds in occupying �e4 h4 23 �c6 Af5+ 24 �d5 f3!
the d5-square, and will therefore stand better than Of course not 24 . . . h3? 2S 4JeS+ and 26 <£\f3.
its opposite number. And second, because the 25 b5 h3 26 � xa7 h2 27 b6 h1� 28
position is open, the bishop is stronger than the �c6 �b1 29 �c5 Jl,e4 White resigned.
knight (although you would not say so, at first
Karpov - A . Sokolov
glance).
Linares cmf(2) 1 987
l ... �f8 2 �fl �e7 3 �e2 �d6 4 �d3
�d5 5 h4Ac8!
After the bishop check at a6, the black king
goes in the opposite direction to the one White 's
7-26
king retreats to .
6 �f3?!
6 f3 Jla6+ 7 �e3 �cS 8 4Jc2 should have
been preferred. Here 's Averbakh 's suggested B?
continuation: 8 . . .Jlfl 9 g3 Aa6 10 4Jd4 Ab7
( 1 0 . . . �b4 1 1 4Jc6+) 1 1 �d3 �b4 1 2 �c2 JldS
13 �b2 g6 14 �c2 a6 l S �b2, and it's still not
clear how Black will break down his opponent's
Which pawn should Black take? In prin­
resistance.
ciple, when you have a knight against a bishop,
6 -'l,a6+ 7 �c3
the task is made easier, the narrower the battle­
•••

On 7 �e3 �cS 8 4::\g S �b4 9 <£\xf7 �xb3,

field: all the pawns should be on the same side.
the a-pawn decides. From this point of view, the logical move is
7 h6 8 �d4 g6 9 �c2?!
•••
1 . . .<£\xa4! And in fact, this would have led to a
9 f3 is stronger, taking the important e4- draw: 2 �d4 �d6 3 �bS (3 eS+ �e6 4 h4 h6 S
square under control. �bS <ticS 6 AxeS be 7 �xcS gS=; 3 �b4 <ticS 4
9 �e4! 10 �e3 f5
•••
�xcS+ be+ S �bS gS=) 3 . . . 4::\ c S 4 AxeS+ be
Black has deployed his king to maximum ( ..6. s . . . gS) s h4 h6 6 �c4 �c6 7 eS hS 0 =.
effect. He intends to drive the knight from e3, 1 � xe4? 2 �b5 �c5 3 .1lf8!
•••

and then to attack the g2-pawn with his bishop. Sokolov probably counted on 3 Ad4? <£\xa4!
11 �d2 f4 12 � g4 4 �xa4 �fS S �e3 �g4 6 �bS �h3= .
Also hopeless is 1 2 <£\c2 ltfl ! 1 3 <tiel �fS 3 . . . � d7

1 29
 Dvoretsky's Endgame Manual

Now 3 . . . 4Jxa4 4 'it>xa4 �f5 5 .lld6+- no In order to understand what follows, we

longer helps. must recall the conclusions we reached when
4Aa3 <it'd5 5 .ile7 <it'd4 6AdS Black studying Rauzer's positions with bishop and
resigned. pawn vs. bishop (Diagrams 4-2 and 4-3). After
8 'it>f6 �xh4 9 'it>e6 �g4 10 �d7 �f5 1 1 �xc8
Tr-auic:;umedies �e6 B lack has no trouble drawing, with the
white pawn on a3 . Here, the pawn is on a2, which
Krnic - Flear would give White a win (although a rather com­
Wijk aan Zee 1 98 8 plicated one), if there weren 't a black g-pawn
on the board. That of course changes the evalu­
ation.
8 . . . 'it>xh4 9 �f5 'it>g3! 10 �e6 �f4 1 1 �d7
7-2 7 �e4 1 2 'it>xc8

w 7-28

A draw was agreed here. Krnic probably just B?

didn't realize that the bishop is completely domi­
nating the knight, and therefore he had every rea­
son to expect a win.
1 <it'f4 �cS ( l . . .'it>f7 2 �e5 or 2 .ll x b6 ab
Doesn't White win now? Not necessarily ­
3 'it>e5 �e7 4 a4) 2 <it'g5 <it'f7 3 a4! a51?
not if his opponent can force the move a2-a3
White has a much simpler task after 3 . . . 4Jb6
and then get back with his king.
4 a5 or 3 . . . a6 4 .llc 5 �g7 5 a5.
1 2 . . . �d3 ! 1 3 'it>d7 'it> c 2 (threate n i n g
4 -'l,c5 (cutting the knight oft) 4 <it'g7 5
••.

1 4 . . . a3!=) 1 4 a 3 �d3 1 5 'it>e6 �e4!

h3!
It turns out White can 't prevent the black
"Steinitz's Rule" in action ! On 5 h4? 'it>f7 6
�h6 �f6, it is White who falls into zugzwang. king from reaching the drawing zone. For ex­
5 <it'f7 6 <it'h6 <it'f6 7 h4 0 (White takes
•••
ample: on 16 Ae7 Black can play either 16 . . . �f4
the opposition, in order to follow up with an out- 1 7 .llf6 �g4! ( 1 7 . . . 'it>e4? is a mistake, in view of
fl anking) 7 <it'f5 (7 . . . �f7 8 �h7 'it>f6 9
•••
18 .ll e 5! with a theoretically won position) 1 8
�g8+- ) s <it'g7 <it'g4 (8 . . . g5 9 h5) 9 <it'f61 �e5 'it>h5 1 9 �f5 �h6 20 Ae5 'it>h7!= (but not
On 9 'it>xg6? �xh4 the black king would get 20 . . . 'it>h5? 21 Ag7+- ), or 16 . . . 'it>d4 17 Ad6
back to the queenside in time: 10 'it>f5 'it>g3 1 1 �c4! (17 . . . �e4? 18 Ae5+- or 18 Ah2+- would
�e6 �f4 1 2 'it>d7 �e5 1 3 �xc8 'it>d5 1 4 .llb6 be a mistake).
�c4=.
Exercises
9 <it' x h4 10 <it'e6 <it'g4 1 1 <it'd7 <it'f5
•••

12 <it' xcS <it'e6 13 <it'c7 <it'd5 14 <it'b6 +-

lt is odd that Flear recommends 3 �h6 (in­
stead of 3 a4) . The GM even awards this move
an exclamation mark, although in point of fact it 7-29
deserves a question mark, and according to analy­
sis by Zviagintsev and Dvoretsky, it probably lets
slip the win.
3 'it>h6? a5! (Flear examines only the weaker
7/8
3 . . . 4Jb6 and 3 . . . a6) 4 Ac5 (4 a4 4Jd6 5 'it>g5 4Jc4)
B?
4 . . . a4! 5 h3 (5 �h7 �f6 6 �g8 g5 is no better)
5 . . . �f6 6 h4 0 �f5 7 �g7 'it>g4 8 �xg6.

1 30
 Bishop vs. Knight

Defensive Methods with a Knight against a Bishop

Sometimes, an inferior position may be saved There is no need to calculate the variation
by tactical mean - using knightforks. But stra­ 7 .. .fg+ 8 'it'xg3 gh+ 9 'itlxh4 Ae2, since the text
tegic methods are also often used. Let's enumer­ provides a much simpler resolution.
ate the most important ones: 8 hg hg 9 gf gf
Blockading the passed pawns; White resigned, in view of 1 0 <£Jd1 ( 1 0 'it'g2
Fixing the enemy pawns on the same eo/or Ae2 1 1 'itlh3 Axf3-+) 10 . . . 'it'e7 1 1 <tlb2 'itld6
squares as his bishop; 1 2 <£Jd1 'itlc5 1 3 <tlb2 'it'b5 14 <£Jd1 'itlb4- + .
Erecting a barrier - the knight and pawns
Nebylitsyn - Galuzin
take control of a complex of important squares,
USSR 1 969
preventing the incursion of the enemy king or at
least making that incursion much more difficult;
Erecting a fortress.
These techniques are not usually employed 7-31
singly, but in combination with each other. How
this plays out, we shall see in the examples from
this section.
B?
Pirrot - Yusupov
Germany tt 1 992

White 's king is tied forever to the king 's

wing. The evaluation of this position hinges on
7-30 whether the knight and pawns can erect an
uncrossable barrier in the path ofthe enemy king.
.l . c4?
.

A tempting, but incorrect pawn sacrifice.

B Black's goal is achieved by l . . .Ad2! 2 <£Jb6 Aa5
3 <£Jc4 Ac7, when there appears to be nothing
that can stop the transfer of the king to a6, fol­
lowed by . . . b7-b5. If 4 <tla3, then 4 . . . 'it'c8 5 <tlb5
1 . . . Axa2 2 'itlf2 t::.. 'itle3 would lead to a Ab8, followed by . . . b7-b6 and . . . 'it'b7-a6-a5 .
roughly equal position. Yusupov finds the best 2 be Ae3 3 a5 Ad2
practical chance. Or 3 . . . 'it'c8 4 <£Jb6+ 'it'c7 5 <£Ja8+ 'it'b8 6
l . . . f5! 2 .£lc3? <£Jb6 Ad2 (6 . . . 'it'a7 7 <£Jc8+ 'itla6 8 <tlxd6; 6 . . . Ac5
His opponent gets greedy : by maintaining 7 <tld7 + 'it'a7 8 <tlxc5 de 9 'itlg 1 'it'a6?? 10 d6+-)
=

the balance of material, he loses the game. 2 <tld2! 7 a6, and we' re back in the game continuation.
was necessary (blockading the passed pawn) 4 a6!!
2 . . . Axa2 3 f4 (fixing the enemy pawn on the After 4 <£Jb6? Axa5 5 <£Ja4 b6 (intending
same color square as the bishop; on the other . . . 'itlc7-b7-a6) 6 <tlb2 'it'c7 7 <tld3 .la.d2 ! (it's im­
hand, 3 'it'f2 f4 4 <£Je4 'itle7 5 'itle1 or 5 g3 was portant to prevent the maneuver <tlcl -b3) White
good, too) 3 . . . 'it'e7 4 'itlf2 .la.d5 6 g3 'it'd6 6 'it'e3 loses.
Ae4 7 'itld4, and there appears to be no way to 4 ba (4 . . . 'it'c8?? 5 a7 Ae3 6 <£Jb6+) 5 -tlb6
•••

break into the fortress White has constructed. Ae3 6 .£l a4 .Q.d4 7 �fl �e7
2 ... d2 (threatening 3 . . . Ae2) 3 �f2 f4! The queenside barrier is erected, and the
White 's pos ition has become hopeless, king can no longer penetrate here. Black there­
since his king is cut off forever from the passed fore tries his last chance : marching his king to
pawn. h4, in an attempt to place his opponent in
4 b3 ab 5 abAd3 6 g3 g5 7 h4 h6! zugzwang. True, White will then play the c4-c5

131
 Dvoretsky's Endgame Manual

break; but then Black can sacrifice his bishop, Exercises

and return his king to the queenside.
8 'llg2 'llf6 9 'llf1 'llg6 10 'llg2 'llh5 11
'll f l 'll h4 12 'llg 2 a5 13 c51 � xc5 14 4) xc5
'llh5
7-34

7-32
7/9
W?

Here 1 5 .£\b7? a4 16 .£\xd6 a3-+ would be a 7-35

mistake. White must set up a fresh barrier.
15 4) a4 'll g 6 16 4)c3 'llg7 17 'll f 1
'llf8 18 'llg 2 'lle7 19 4) b51 a4 20 'llf 1 'll d 8
21 'llg 2 7/ 1 0
Draw, since Black's king can advance no W?
further.

Balashov - Smyslov
Tilburg 1 977

7-33

B?

The game continuation was: 1 �a1?? 2•.•

g3 ( D. 3 h4), and White won easily.

Black draws after l . . .�xe3! 2 fe .£\f8! . The
knight moves inexorably to c5, from where it
deprives the enemy king of the important squares
d3 and e4 (barrier); after this, Black plays . . . �g7-
f6 and . . . h7-h6. 3 �f2 .£\d7 4 �f3 .£\c5 5 �g4
�g6 is not dangerous, since there can be no
bishop check from d3 . If White 's king heads for
the b-file, Black defends the knight with his king
from d6.

1 32
 Bishop vs. Knight

The Knight is Superior to the Bishop

Domination and Knight Forks A picturesque domination of the knight over

the bishop ! Note that the knight takes away only
Nepomniaschy - Polovodin three of the bishop 's squares. Another is con­
Leningrad eh 1 98 8 trolled by the d5-pawn (the pawns' placement
on the squares of the same color as the opposing
bishop is one of the means of restricting its mo­
bility). But the chief blame for White 's helpless­
7-36 ness lies with his own kingside pawns, placed
on squares the same col or as his bishop, and turn­
ing it "bad."
1 1 �c3 �a3 0 White resigned.
B?
R. Reti, 1922

l �c3!
•••

7-38
The key to the position is that on 2 a3 4Je2!!
decides: 3 ab (3 '<l9xe2 ba ) 3 ... 4Jf4+. Without
this little combination, based upon a knight fork,
there would be no win (with the bishop at f3 ,
W?
let's say, the position i s drawn) .
2 .Q.f3 � x a2 3 .Q.dl b3!
Once again, Black has recourse to a fork,
in order to advance his passed pawn ( 4 Axb3
l �d4+! �c5
4Jcl + 5 '<l9c3 4Jxb3 6 'it>xb3 '<l9b5 0 loses at once).
On l . . .'<l9b7 2 '<l9xh2 '<l9a6 3 4Jb3 ..11f4+ 4
On the other hand, 3 . . . 'it>b5 4 -'i.c2 4Jc1 + 5 'it>d2
'<l9h3 'it>b5 5 '<l9g4 -'i.b8 6 f4 '<l9b4 7 f5 (7 4Jd4?
b3 is strong, too.
..11 xf4) 7 . . . '<l9xb3 8 f6 '<l9b4 9 f7 ..11d6 10 a6+­
4 �d2 b2 5 .Q.c2 (5 'it>c2 4Jc3) 5 �b4 .•.

(pants) is decisive. In this line, B lack needs just

6 .Q.bl �b6!
one tempo; therefore, with the king already at
An outstanding loss of tempo ! The straight­
c5, this line would not work: 2 4Jb3+? '<l9b5 3
forward 6 . . . '<l9b5? 7 ®c3 0 4Jc6 (7 . . . ®a4 8
'<l9xh2 ..ltf4+ 4 '<l9h3 ..ltb8 5 '<l9g4 '<l9b4 = , or 2
'it>xb2=) 8 -'i.a2 4Je7 9 '<l9b3 leads to a draw.
'<l9xh2? .llf4+ 3 '<l9h3 '<l9xd4 4 a6 �b8=.
7 �c3 �b5 0 8 �b3 (8 '<l9xb2 ®c4-+ )
2 �ht n o +-
8 �c6 9 �c3
There is not one square for the bishop where
..•

9 -'td3 + ®a5 1 0 ® x b 2 ®b4-+ is no

it would not be vulnerable to a knight fork.
better.
9 � a 4 1 0 � x b 2 ( 1 0 -'i. a 2 '<l9 a 3 )
Exercises
••.

1 0 � b4! 0 (but not 1 0 . . . 4Jxd4? 1 1 -'i.a2)

...

7-3 7 7-39

w 7/ 1 1
W?

1 33
 Dvoretsky's Endgame Manual

Exercises

7-40 7-42

7/ 1 2 7/ 1 3
W? B?

Fixing the Pawns Closed Position, Bad Bishop

We have already pointed out more than once In positions with pawn chains, the bishop
how important it is to fix the enemy pawns on has limited mobility, and therefore is sometimes
the same color squares as his bishop. Thus, weaker than the knight. The chief reason for a
we limit ourselves here to looking at two new bishop being "bad" is that his own pawns are
examples. fixed on the same color squares as the bishop.

Osnos - Bukhman Zubarev - Alexandrov

Leningrad eh tt 1 968 Moscow 1 9 1 5

7-43
7-41

w
w

The exploitation o f the knight's indisput­

The advantage, of course, is White 's, since able advantage over the bishop is uncomplicated,
all his opponent's pawns are isolated and weak. but quite instructive. First and foremost, the king
But this might not have been enough to win, had must be made as active as possible, and there 's
White not found the following maneuver, to force an open road for him straight to c5.
the d-pawn to advance onto a square the same
1 �f2 �e7
color as his bishop.
On l . . .�f7 White neutralizes his opponent's
1 4Jd5 �g7 2 4)b6! -'lb3 3 4)c8! d5 4
activity on the kingside by erecting a barrier: 2
.£\e7 h5 5 h4!
�e2! �g6 3 �e3! �gS 4 g3! , after which the
One more pawn fixed on a light square. king continues its march to c5.
5 �f8 6 4Jf5 -'tc2 7 .£\e3 -'lb3 8 �h2
2 �e3 �d8 3 �d4 �c7 4 �c5 jlc8
••.

�e7 9 4Jf5+ �e6?? (a terrible blunder in a

The next phase flows from the two-weak­
hopeless position) 10 4)d4+ B lack resigned.
nesses principle. White cannot yet win on the
queenside alone; therefore he sends the knight
(via f4) to the kingside, to harry the enemy
pawns. These in turn will have to be advanced,
which will make them much weaker than they
are in their initial positions.

1 34
 Bishop vs. Knight

5 .£l b4Ab7 6 g3 tZ:l xg4!? gh 4 'it'g2. And 2 'it>g2!? gh 3 g4! is also

It 's useful to deprive the opponent of tac­ worth looking into.
tical chances (such as . . . d5-d4). 1 . . . gh?!
6 ...Ac8 7 .£ld3 Ad7 8 .£lf4 ( £:. 9 4Jh5) After home analysis, Kasparov decided to
8 ... g6 9 .£lh3! ( £:. 10 4Jg5) 9 . . . h6 10 .£)f4 g5 alter the pawn structure, judging (correctly) that
11 .£lh5 Ae8 12 .£lf6 Af7 13 .£l g4! after 2 gh .llg6 White could no longer break
One more black pawn must now be moved through. Alas, neither he nor his trainers could
to the same color square as its bishop. foresee White 's tremendous retort, securing his
13 ... h5 14 .£)e3 Ag6 (14 . . . h4 15 gh gh king a road into the enemy camp.
1 6 4Jg2 ; 1 4 . . . g4 1 5 4Jg2 and 1 6 4Jf4) 15 h4! 2 .£)g2!! hg+ (2 . . . h3 3 4Jf4) 3 � xg3 �e6
Fixing the pawns! 4 .£)f4+ �f5 5 .£) xh5 (threatening 4Jg7-e8-
15 . . . gh 16 gh ( £:. 4Jg2-f4) 16 . . Ae4
. 17 c7) 5 . . . �e6 6 .£)f4+ �d6 7 �g4 Ac2 8
.£1ft Af3 18 .£ld2 Ae2 19 .£l b3 Ag4 20 �h5 Ad1 9 �g6
.£)d4 0
The concluding phase of White 's plan is
to create a zugzwangposition. For this the knight
needs to be brought to f4, tying the bishop to the
7-45
defense of two pawns at once.
20 . . .Ah3 21 .£le2 Af5 22 .£l f4 A g4
23 b4 0
The end is achieved !
B?
23 . . . �d7 24 �b6 Af3 25 � x a6 �c6
26 .£) xe6 Black resigned.

Karpov - Kasparov
Moscow wm (9) 1 984/85 9 ... �e7!
9 . . . .1lxf3 10 'it>xf6 is absolutely hopeless. In
such situations, we employ the steady driving
offofthe enemy king: the knight goes to f5 , and
7-44 after the king 's forced retreat (since the pawn
endgame is lost), White's king goes to e5 or e7.
Then the knight gives check again, etc.
10 .£) xd5+?
B? Unjustified greed - now B lack gets the
chance to activate his king, via the newly-opened
d5-square.
1 0 4Jh5! .ll xf3 1 1 4Jxf6 was far stronger, for
White's task here is considerably more com­ instance : 1 l . . . 'it>e6? 1 2 4Je8 ( £:. 1 3 4Jc7+)
plicated. For the time being, the king has no route 1 2 ... 'it>d7 (12 ... .1le4+ 1 3 'it'g5 'it>d7 1 4 4Jf6+ 'it>e6
into the enemy camp; he must continue by "wid­ 1 5 4Jxe4 de 16 'it'f4 'it>d5 17 'it>e3 0 +- ) 13 4Jg7
ening the beachhead. " The interfering kingside 'it>e7 (otherwise 14 'it>f6) 14 'it>f5, and White wins.
pawns can be removed in two ways: by g3-g4, The best defense would be: 1 1 . . . .1le4+! 1 2
or by exchanging on g5, followed by f3-f4. 'it>g5 .ild3!.
The best defense was 1 . . .'it>e6! . On 2 hg fg
3 f4, Black can draw either by 3 . . . gf 4 gf A g6,
or by 3 . . . g4!?. And after 2 g4 hg 3 hg, as John
Nunn points out, Black must play 3 . . . gf! (3 .. .fg
4 tZ:l xg4, followed by 'it>g3 and f3-f4, would be
weaker) 4 \t>xf3 (4 gf A e4) 4 . . .fg 5 'it>g4 'it>f6 6
CiJ x d 5 + 'it>g6 = . White keeps more practical
chances by refraining from 3 hg in favor of 3

135
 Dvoretsky's Endgame Manual

And White will not have an easy trip back:

on 'it>h6 there follows . . . 'it>f6, and Black's king
advances. White should continue 22 �b8 Ac2
23 �c6+ 'it>e6, when he has two ways to reach
7-46
the goal :
a) 24 �f8 Ag6 25 �a7 Ad3 26 'it>e8 Ae2
(26 . . .'it>f5 27 �d7 'it>e4 28 �c6 +- ) 27 �c6 'it>d6
28 �e7! (28 �e5 �e6 29 'it>d8 'it>f5 30 �c6 'it>e4
W?
31 'it>c7 'it>d3! 32 'it>b6 'it>c4 would be weaker)
28 . . . 'it>e6 29 �d8 �d6 (29 . . . -ilfl 30 �c6) 30
�f5+ 'it>e6 3 1 �e3 �d6 32 'it>c8 'it>c6 33 �b8 +- ;
b) 24 �e5 'it>f5 (after the passive 24 . . . �e7
White can't gain control of the f6- or e5- 25 �g4, White brings the knight to e3, then re­
squares with his king. Capturing the pawn is turns the king unhindered to its own side, and
also unconvincing : 1 3 �xd5+ �d6 1 4 �c3 ( 1 4 begins preparations for a3-a4, bringing the knight
� e 3 o r 1 4 � f4 d o n ' t c h a n g e anyth i n g ) to c3 at the right moment) 25 'it>f7 'it>e4 (25 . . . Ad1
1 4 . . . �fl ! 1 5 �f4 �g2 1 6 �e3 (intending �e4- 26 �c6! �e4 27 'it>e6) 26 �c6 Ad1 27 �e6 (27
c5) 16 . . . Ah3! 17 �e4+ �d5 18 �c5 Ac8 19 'it>d3 'it>e7? �d3 28 'it>d6 �c4 29 �e5 Af3 would be
Af5 + 20 'it>c3 Ac8, and Black is apparently out inexact) 27 . . ..1lh5 28 �d7! (but not 28 'it>d6? Ae8
of danger. 29 � e 7 'it> x d4 30 � x d 5 'it> c 4 = ) 28 . . . Ag6
I thought that the variations I had found (28 . . . Ae2 29 'it>d6 and 30 'it>c5 ; 28 . . . �d3 29 a4!)
were sufficient to demonstrate the position was 29 a4! ba 30 b5 a3 31 �b4 +- .
drawn. However, grandmaster Mihail Marin sug­ 10 Citle6
•••

gested an extremely dangerous plan: 13 �g4! 10 . . . 'it>d6!? was more exact, leading, after
with the idea of continuing �e5-c6-b8. 1 1 �c3 (or 1 1 �xf6 Axf3) 1 l . . .Axf3 12 'it>xf6
I attempted to hold the line by 13 . . . �f1 ! 1 4 .ilg2, to a position examined in the last note.
� e 5 Ah3, and now 1 5 �c6+ 'it>d6 1 6 �a5 ( 1 6 1 1 4:)c7+ Citld7?
�b8?? �c8 and 1 7 . . . �c7) 1 6 . . . 'it>e7! 1 7 �b3 'it>f7 Now B l ack w i l l be two pawns down .
18 �c5 Ac8 is useless; while winning the d-pawn 1 1 . . . �d6 was stronger. I f 1 2 � e8 + , then
by 1 5 �g6+ 'it>f7! 16 �f4 �c8 17 �xd5 'it>e6 12 . . . �e7 (12 ... �d5 1 3 f4 is inferior) 1 3 �xf6
would lead to the drawn position we already �xf3 14 'it>f5 'it>d6 1 5 'it>f4 Ag2 16 'it>e3 Ah3,
know. But Marin showed that White could play leading to roughly the same positions as after
for zugzwang : 1 5 'it>g6! 'it>e6!? 1 6 �c6 'it>d6 1 0 . . . �d6. And on 1 2 � x a 6 there fo l l o w s
( 1 6 . . . Af5+ 17 'it>g5 and 18 �b8) 17 �a5 'it>e7 1 2 . . . Axf3 1 3 �xf6 �d5 .
18 �b3 Ad7 19 �c5 �c8 20 'it>g7 0 (but not 20 12 4:) xa6 -'lxf3 13 Citl xf6 Cit'd6 14 Cit'f5
'it>g5 'it>f7) - Black loses the a6-pawn. Cit'd5 1 5 Cit'f4 -'lh1 16 Cit'e3 Cit'c4 17 4:)c5
On the other hand, Black's resistance is not .1lc6 18 "il d3 -'lg2
yet broken - he can lock the king in at g7 for a 18 . . . Ae8!? 1 9 �e5+ 'it>d5 was worth con­
while by 20 . . . Af5 2 1 �xa6 �d3 . sidering. Even with two extra pawns, the out­
come is still far from clear - Black's king is too
active. He must only be careful not to go after
the a3 -pawn (when White will lock him in by
7-4 7
putting his own king at c3).
19 .'ile5+ Cit'c3 ( 1 9 . . . �d5!?) 20 4:)g6 Cit'c4
21 4:)e7

1 36
 Bishop vs. Knight

One interesting try is 25 'itt c 3 <ittd 6! (other­

wise 26 �b3 followed by 27 a4 - Black can 't
reply 26 . . . Ac6, because of 27 d5+!) 26 a4 (26
®b3 Ac6 27 d5 .ile8 28 'itt c3 'itte 5=) 26 . . . ba 27
7-48
<tlc4+ <itfd5 ! (27 . . . 'itt c7? 28 b5 +- ) 28 <tlb6+ (28
b5 Ae2 29 b6 <ittc6 30 d5+ �b7 31 d6 <ittc 6=)
28 . . . ®c6 29 <tlxa4 <itfb5 (or 29 . . . �h5). Paradoxi­
cally, two extra pawns are insufficient to win here
B?
- White has no way to strengthen his position.
And nevertheless, Karsten Muller has found
a subtle means of getting the knight to the key
square c3 . After 22 . . . 'itt d 5! 23 'ittd 3 <itte 6! , White
2 t Ab7?
•.. plays 24 4Jg7+!! 'itt d7 (24 . . . �d6 25 4Je8+) 25
2 l . . .'itfb3? would not have worked in view 4Jh5 . On 25 . . . �d6, there follows, not 26 <tlf4?
of 22 d5 <itfxa3 23 d6 .ilh3 24 <tld5 . However, it Ac8! and 27 . . . �f5+, but instead 26 <tlg3(f6)! , and
would be safer to keep the bishop in the lower then 27 <tle4+ and 28 <tlc3. And on 25 . . . �g2 (hop­
half of the board: 2 1 . . .�h1 ! 22 <tlf5 (22 d5? ing fo r 26 <tl g 3 ? � e 6 ! 27 <tle4 �fl + and
.ilxd5=) 22 . . . <itfd5. Many analysts have diligently 28 ... 'ittd 5), then 26 4Jf4! Afl + 27 �e4 �d6 28
examined this position, but none have been able 'itte 3! (zugzwang) 28 ... Ac4 (28 ... �c6 29 d5+ <ittd6
to find a win here. The move Black actually 30 <itfd4 Ac4 3 1 a4 +- ) 29 4Je2 +- (29 . . . <itt d 5 30
played is a decisive mistake. <tlc3+, when the c4-square, which is needed by
22 4)fS Ag 2? the king , is occupied by the bishop) .
As Speelman and Tisdall indicated, nei­ 23 4)d6+ \tlb3 24 4) x bS \tla4 25 4)d6
ther 2 2 . . . �c3? 23 'itt f4! 'itt b 3 24 <tle7 �xa3 25 Black resigned.
d5, nor 2 2 . . . .ilc6? 23 <itff4 'itt b 3 24 'itt e 5 <itfxa3
25 <itt d 6 .11e 4 26 <tlg3 would save Black. He Exercises
had to play 2 2 . . . <itfd5 ! 2 3 'itt d 3 <itt e 6! . For ex­
ample: 24 <tle3 (24 <tlg3 �g2 25 <tle4 .llfl + 26
'itt e 3 'itt d 5 27 <tlc3 + 'itt c 4) 24 . . . .ilf3 ! (it's im­
portant to prevent White 's knight from reach­
ing c 3 ) .
7-50

7-49 7/ 1 4
W?

2 5 d 5 + 'itt e 5 ! (but not 2 5 . . . �xd5? 26 �d4 7-51

D. 27 'itt c 5) leads to nothing. On 25 'itt d 2, in­
tending to march the king into the enemy camp,
B lack responds, not with 25 . . . <itt f6? 26 <tld1 !
and 2 7 <tl c 3 ( s i n c e B l a c k n o l o n g e r h a s 7/ 1 5
26 . . <itfd5),but simply waits - when White's king
.
W?
reaches the 8th rank, the bishop will cut off its
path to the queenside along the h3-c8 diagonal.

137
 Dvoretsky's Endgame Manual

Chapter 8

Rook versus Pawns

Practically all these endings are "rapid"; the mastering them does not free us from the neces­
outcome of the fight depends, as a rule, on a sity of deep and precise calculations, but makes
single tempo. We shall study typical techniques; this job much easier.

Rook vs. Pawn

"Moving Downstairs " When the black pawn reaches a3 it will be

abolished by means oH'lg3 (the pawn may come
First let us look at the rarest case, when a even to a2 and then perish after :8gl followed
pawn is stronger than a rook. by :8 a l ) .
With Black o n move, after 1 . . .\t'bS(cS)! the
G Barbier, F. Saavedra, 1895 position is drawn, because cutting the king off
along the 4th rank brings nothing.
In the starting position, let us move the black
king to c6 and the pawn to bS. The strongest
8- 1 move is still l :8g5!, but Black can respond with
l . . .'�b6. However the king transfer to the a-file
loses time, and its position is less favorable there
than on the c-file (where it "gives a shoulder
W kick" to the rival king). After 2 \t'g7 �aS 3 \t'f6
White arrives in proper time to stop the pawn.

Pawn Promotion to a Knight

1 c7 E! d 6 + 2 ® b 5 ! ( 2 � c 5 ? :8 d l )
2 . . . E!d5+ 3 ®b4 E!d4+ 4 ®b3 E!d3+ 5 ®c2
This maneuver, which helps the king to
avoid checks, is what we call "moving down- 8-3
stairs . " However the fight is not over for the
moment.
5 E!d4!
. ••

I f 6 c8�? then 6 . . . :8 c4+! 7 �xc4 stalemate. W

6 c8E!!! ( �::,. 7 :8 a8+) 6 E!a4 7 ®b3! + -
•••

Cutting the King Off

1 E!h2+ ®cl 2 ®c3 b l .l£\ + ! 3 ®d3
.l£\a3 4 E!a2 .1£\bl! leads to a draw.
It is worth mentioning that the erroneous
8-2 4 . . . <£\bS? loses the knight. In rook-versus-knight
endings, one should not separate the knight
from the king.
Black can also save himself by stalemate :
l . . .�bl ! 2 �b3 �a l ! 3 :8 xb2. However, with a
bishop or a central pawn his only drawing pos­
sibility is pawn-to-knight promotion.
If he has a rook pawn instead, this method
1 E!g5! +- does not work.

138
 Rook vs. Pawns

<it'a2 5 .§ b7 0 is hopeless. Correct is 1 �b2!

••.

2 .§b8+ (2 .§h2+ �b3! , rather than 2 . . . <it>bl? 3

<itJc3) 2 �c1! 3 .§aS �b2 4 �d2 a2 5 .§b8+
8-4
••.

�a1 ! .

An Intermediate Check for a Gain

w of Tempo

Korchnoi - Kengis
Bern 1 996
1 �c4 a2 2 �b3 a1� + 3 �c3 0 +­
By the way, an additional pawn at b5 could
not have helped Black.
8- 7

8-5
B

w
Kengis resigned in this position, depriving
his opponent of the opportunity to demonstrate
an exemplary winning solution :
1 �b4 a2 2 �b3 a 1 � + 3 �c3 b4+ 4 1 �f2 2 .§f8+!
.••

� x b4 �c2+ 5 �c3 �e3 6 .§h4! (another 2 <itJd3? g3 3 .§f8+ <it>el ! leads only to a draw.
option is 6 <it>d3 .!lJd5 7 .§ h4 �b2 8 .§ d4 and the 2 �e2 3 .§ g8! �f3
.••

knight, being separated from the king, will die Because of the intermediate check, White
soon) 6 �a2 (6 . . . .!lJdl + 7 <it'd2 .!lJb2 8 .§ b4
•••
succeeded in driving the opposite king back one
<it>a2 9 <it>c2 <it>al 10 .§ b8; 6 . . . .!lJd5+ 7 �b3 <it'cl square, from f2 to f3 .
8 .§ c4+ �bl 9 .§ d4) 7 .§a4+ �b1 8 .§e4 �f5 4 �d3 g3 5 .§f8+ �g2 6 �e2 + - .
9 .§e5 �d6 10 �b3 �c1 1 1 .§c5+ �b1 1 2
.§d5 +- . Shouldering

Stalemate

We have already seen a case of stalemate 8-8

that has practical value (diagram 8-3). The fol­
lowing position is also worth keeping in mind.

8-6

1 .§h2+ �a3!
Black achieves a draw by not allowing the
B? white king to approach the pawn. l . . . � b l ? is
erroneous in view of 2 'it>b3 al.!lJ+ 3 <itJc3 .

Let us look at a slightly more complicated

l . . .a2? 2 .§b8+ �a3 3 �c2 ! al.!lJ+ 4 �c3 case in the following diagram.

139
 Dvoretsky's Endgame Manual

I. Maizelis, 1950 T.-aulwmedie§

Neumann - Steinitz
Baden-Baden 1 870
8-9

8-11

B?

l . . .a5? is bad because of 2 § h5! (cutting

the king oft). However l . . St'b5? 2 �f7 a5 3 �e6
a4 4 �d5 is no better. 1 �f8
Only 1 ... �c5! holds. Black does not al­ The simplest way is 1 �g8! �g6 2 �h8=.
low the white king to approach his pawn. 1 ... �f6 2 g84) +! �e6 3 4) h6 §h7 4
4) g4??
Outflanking As we already know, after 4 <tlg8! the game
would have been drawn. Now White is lost.
Shouldering and outflanking ideas are dis­ 4 . . . §h4
tinctly repre sented in the following famous 4 . . . §h3! could have won immediately.
endgame study. 5 4)e3 (5 <tlf2 § f4+) 5 . . . §e4 6 4) d 1
§f4+ 7 �g7 §f3 8 �g 6
8 <tlb2 �d5 9 <t!a4 §b3 !::.. 1 0 . . . �d4 and
R. Reti, 1928
1 l . . .§b4 makes no difference.
8 ... �e5 9 �g 5 �d4 10 �g4 §f1 1 1
4)b2 §b1 1 2 4) a4 §b4 White resigned.

8-1 0
Fries-Nielsen - Plachetka
Rimavska Sobota 1 99 1

W?
8-12

1 §d2(d3)!! d4 2 §d1! �d5 3 �d7!

Black is in zugzwang : if 3 . . . �c4, then 4 W?
�e6 and if 3 . . .'�e4, 4 �c6.
1 § d 1 ? is erroneous: l . . .d4 2 �d7 (2 �f7
�e4 3 �e6 d3) 2 . . . �d5 ! (Black prevents an out­
The actual continuation was 1 �c6? h5
flanking) 3 �c7 �c5 ! (3 . . . �c4? 4 �d6! d3 5
=

2 �d5 h4 3 �e4 h3 4 �e3 �g 3 5 §e1 h2

� e 5 ) , and it is White who has fallen into
6 �e2 �g 2 7 §h1 � xh1 8 �fl Draw.
zugzwang.
1 § e8? is no better: l . . .h5 2 § g8+ �f3 3
§h8 �g4 4 �c6 h4 5 �d5 h3 6 �e4 �g3 7
�e3 �g2! (rather than 7 . . . h2?? 8 § g8+ �h3 9
�f2 ! h 1 <tl + 1 0 �f3 �h2 1 1 § g7 0 ) 8 § g8+
�fl != or 8 �e2 h2 9 § g8+ �h1 !=.
White should have gained a tempo by means
of the intermediate check: 1 § g2+! �f4 (after

1 40
 Rook vs. Pawns

1 . . . �h3 2 � g8 h5 3 �c6 h4 4 �d5+- the black

king, pressed to the edge of the board, is placed
extremely badly) 2 �h2! �g5 3 �c6 h5 4 �d5
8-1 6
h4 5 �e4 �g4 6 �g2+ �h3 7 �g8+ - .

Alekhine - Bogolj ubow

Germany!The Netherlands wm (19) 1929
8-3
W?

8-13

B? 8-1 7

The world championship challenger played 8-4

l. .. �g4?? and resigned after 2 b7 fS 3 b8� W?
.§. x b8 4 .§. xb8 f4 5 �dS f3 6 �e4 f2 7 .§.f8
�g3 8 �e3
He should have applied the shouldering
method: l . . .�e4!. It is easy to see that in this
case the position would have been drawish: the
black king prevents his opponent from getting
to the black pawn in time. 8-18

Exercises

8-5
W?
8-14

8- 1
W? 8-1 9

8-6
W?/Play
8-15

8-2
W?

141
 Dvoretsky's Endgame Manual

Rook vs. Connected Pawns

If two black pawns are placed on the 3rd If 2 a7? now, then 2 . . . �c3 3 'iffb7 �b4 4
rank, or one pawn has reached the 2nd rank b6 'iff b 5 = . Here we observe "th e tail-hook"
while the other is on the 4th rank, a rook can­ again; the techniques that we know from bishop
not stop them. Sometimes, however, White can versus pawn endings (diagram 4-29).
save himself by creating checkmate threats, 2 ®a7!
when the black king is pressed to an edge of the Black resigned in view of 2 . . . �c3 3 b6 �c4
board. 4 b7 l:'! b 1 5 b8� l:'! xb8 6 �xb8.
We call this method "a change of th e
B. Horwitz, J. Kling, 1851 leader." Why does White push the less advanced
b - pawn? F irst of all, because the rook, being
placed on another file, does not prevent its march.
In addition to it, the a-pawn that remains on the
8-20 board after gaining the rook is more remote from
the black king, so its "tail holding" will be more
difficult.

w In a battle against two connected passed

pawns, the best position for the rook is behind
the more advanced pawn.

1 ®f5 ®h4 2 ®f4 ®h3 3 ®f3 ®h2 4

®e3! <if} g 2
Or 4 . . . �g3 5 l:'!g1 + 'iffh 4 6 �f4 'iff h 3 7 'ifff3, 8-22
and here 7 . . . \fth2?? 8 l:'! b 1 even loses for Black
in view of zugzwang.
5 ®d3 ®f3 6 ®c3 a2 7 <if} xb2 (or 7
l:'! fl +) with a draw. W?

The fo llowing simple examp le demon­

strates several very important practical ideas.
1 §. g6! ®d7 2 § g4! g 2! 3 §. x g 2 <if}e6 4
Topalov - Beliavsky §. g 5! and White wins because the black king is
Linares 1 995 cut off from the pawn along the 5th rank.
Sozin demonstrated a similar position in
1 93 1 , with the only difference that the white king
stood on a7 . In that case, after 1 l:'! g6! 'iffd7 an
8-21 alternative solution occurs : 2 'it1b6 �e7 3 'it1c5
'ifft7 4 l:'! g4 �f6 5 'iffd 4! (5 l:'! xf4+? �g5 6 l:'! f8
'iffg4 7 �d4 g2=) 5 . . . \ftf5 6 l:'! g8 +- .
This line does not work when the king is
W? placed on a8: 1 l:'! g6! �d7 2 �b7? 'ifte7 3 �c6
'ifft7 4 l:'! g4 �f6 5 �d5 �f5 6 l:'! g8 f3! 7 ®d4 (7
l:'! xg3 �f4 8 l:'! g8 f2= ; 7 l:'! f8+ �g4 8 �e4 f2 9
�e3 �h3=) 7 . . .f2 8 ®e3 f1 �+! with a draw.
After 1 b6?, l . . . l:'! xa6? 2 b7 l:'!a5+ 3 �c4
etc., loses (moving downstairs). Black holds with It should be noticed that the rook should be
the intermediate check prior to the capture of placed in the rear of the more advanced pawn
the pawn: l . . . l:'! a 5 + ! = . similarly, even when other forces conduct the
1 <if}b6 <if}d2 fight.

1 42
 Rook vs. Pawns

Alekhine - Tartakower adj ourned position was drawish, White could

Vienna 1 922 have held it by means of checkmate threats :
1 �d6! �c8 (1 . . . �e8 2 ®e6 �f8 3 �f6=)
2 Etc1 + �b7 3 Etb1 + �a6 4 �c6 �a5 5
�c5 �a4 6 �c4 �a3 7 �c3 �a2 8 .§.f1
8-23 h5 9 �d3 = L::. 10 �e3 ; 10 � xf2.

Fridstein - Lutikov
USSR eh tt, Riga 1 954
W?

8-25
Alekhine analyzes the natural continuations
1 ®c2, 1 �c4, 1 g5, 1 �h2 and shows that all of
them are good enough at best for a draw. But his
W?
beautiful concept wins:
1 Etd5!!
"The variations springing from this rather
unlikely move (it attacks one solidly defended
Another case of a totally groundless ca­
pawn and allows the immediate advance of the
pitulation. The intermediate check 1 .§.b4+! led
other) are quite simple when we have descried
to a draw.
the basic idea - the black pawns are inoffensive:
I ) When they occupy squares of the same eo/or
Maroczy - Tarrasch
as their bishop, for in that case White's king can
San Sebastian 1 9 1 1
hold them back without difficulty, by occupying
the appropriate white squares, e.g. l . . .f2 2 �d1
e4 3 �c2 .llf4 4 � fl followed by 5 �d1 ; and 2)
When the rook can be posted behind them, but 8-26
without loss of time, e.g. l . . .e4 2 � f5 .llg 3 3 g5
e3 4 � xf3 e2 5 � e3" (Alekhine).

Arulaid - Gurgenidze
Lugansk tt 1 956
After 1 � xh2! �xh2, an immediate "change
of the leader" wins : 2 ®a6! ®g3 3 b5 ®f4 4 b6
�e5 5 b7 � b 1 6 �a7 ®d6 7 b8�+ � xb8 8
8-24 �xb8 +- . The move 2 a6? misses the w i n :
2 . . . �g3 3 �b6 �f4 4 b 5 �e5 5 �a7 (5 a 7 �d5
6 �b7 �c5=) 5 . . . ®d6 6 b6 �b1 ! 7 �b7 (7 b7
®c7) 7 . . . ®c5=.
w White could also have played 1 ®a6! � a4
( l . . .h1� 2 � xh1 .B: xh1 3 b5) 2 .B: xh2 .B: xb4 3
.B:h5 1:::,. 4 .B:b5 +- .
The actual continuation was 1 �c6??
The game was adj ourned and White re­ I;tc1+ 2 �b6 I;tc4! ( L::. 3 . . . .B: h4) 3 I;t x h2
signed without resuming the play. However the I;t x b4+ 4 �c5 I;ta4 5 �b5 I;t xa5+ Draw.

1 43
 Dvoretsky's Endgame Manual

Penrose - Perkins The opponent's arguments and the author­

Great Britain eh, Brighton 1 972 ity of the book convinced Tseshkovsky, and he
accepted the proposed draw.
It was however an unfounded decision !
B lack 's play can be improved by means of
8-2 7 3 . . . <;!;>c5! (instead of 3 . . . <;!;>c4?) 4 � c8+ (4 <;!;>e4
b2 5 �c8+ ®d6 -+ ) 4 . . . ®d4 5 � d8+ <;!;>e3 6 �b8
b2. Curiously enough, the resulting position is
examined on the same page of the same book
B? and, as Tarrasch proved in 1 9 1 2, it is won !
7 ®e5 ®f3! (rather than 7 . . . c2? 8 �b3+) 8
®f5 (8 �b3 'ittg4 -+ ) 8 . . . ®e2 ! 9 ®e4 ®dl 1 0
®d3 c 2 1 1 � h8 c l .f:l + ! and 1 2 . . . bl � -
This position is evaluated as drawn in
Encyclopaedia of Chess Endings . In fact B lack Exercises
can win it rather simply by means of shoulder­
ing followed by moving downstairs.
1 . . .'�1e4! 2 � g4+ (2 �g7 <;f;>f3 -+ ) 2 . . . <;!;>f3 3
� xh4 g2 4 �h3+ <;f;>f4 5 � h4+ ®f5 6 � h5+ <;f;>f6
8-29
7 � h6+ <;!;>g7 -+ .
The game continued l �f4? 2 �d4 . . .

�f3 (2 . . . h3 3 �f8+ <;!;>g4 4 <;!;>e4 h2 5 � g8+ <;!;>h3

6 <;f;>f4=) 3 E!f8+ �g 2 4 �e3 h3 5 E!h8 �h2
8-7
(5 . . . h2 6 <;f;>f4=) 6 E! g8! g 2 7 �f2 �h1 8 E! g7
W?
h2 9 E! x g 2 Draw.

A. Petrosian - Tseshkovsky
USSR eh ( I ), Minsk 1 976

8-30
8-28

8-8
B? W?

B lack has an elementary win : l . . .®d7!

(threatening with 2 . . . c2 or 2 . . . b3) 2 � a7+ <;!;>d6
3 � a6+ ®d5 , etc. He played less precisely:
l �d5?! 2 �f5
...
8-31
In this position, the game was adjourned.
Later in a hotel room, Petrosian demonstrated
the following continuation to his rival : 2 . . . b3
(2 . . . c2 3 � dB+ ®c4 4 <;!;>e4!=) 3 � d8+ <;!;>c4 4 8-9
<;!;>e4, and showed him a volume of Chess End­ W?
ings edited by Averbakh where the final posi­
tion is evaluated as drawn in connection with
the line 4 . . . b2 5 �c8+ 'it>b3 6 �b8+ ®c2 7 <;!;>d4=.

1 44
 Rook vs. Pawns

Rook vs. Separated Pawns

Exercises

8-32

8-34

8- 1 0
W?
If four files separate the pawns, then the
rook can stop them without help of its king.
1 § b l ! ( p arry ing the threat 1 . . . 'it'f2)
l. .. <i!}d3 ( .6. 2 . . . ®c2) 2 §gl l =
Move the b2-pawn to c2 . Now the position
is lost (1 .§. c l ®d2 -+ ).
8-35

8-33
8- 1 1
W?

l ... c4 2 <i!lf5 8-36

2 fl. c8 c3!, and if 3 fl. xc3, then . . . ®b2 and
the a-pawn promotes.
2 c3 3 §cS
•••

After 3 ®e4 c2 4 fl. c8 'it'b2 5 fl. b8+ �c3 6 8- 1 2

.§. c8+ both 6 . . . 'it'd2 and 6 . . . 'it'b4 win. W?
3 <i!}b2 4 §bS+ <i!}c2(a3) 5 §aS <i!}b3
•••

6 §bS+ <i!}c4 7 §aS c2 - + .

This is perhaps all one should remember

about this sort of position. Some additional
ideas are shown to you in the exercises for this
section. 8-3 7

8- 1 3
W?

1 45
 Dvoretsky 's Endgame Manual

Chapter 9

Rook Endgames

Rook endings are perhaps the most impor­ very limited number of precise positions; as it is
tant and most difficult kind of endgame. Most highly improbable that one would meet them in
important, because they occur in practice much practical play. So, mastering the basic ideas and
more often than other endings. Most difficult, methods is fully sufficient in those cases.
because one must absorb and remember a much In rook endings, however, a sophisticated
greater volume ofknowledge than in endings with theory of positions with reduced material exists
other material relationships. (for example, those with R+P against R), and these
The reason is that, in other endgames, situ­ situations occur very often in practice. This means
ations with a minimum number of pawns on the that we cannot omit studying a considerable num­
board are either elementary or not very impor­ ber of precise positions.
tant. Therefore one needs only to remember a

Rook and Pawn vs. Rook

The Pawn on the 7th Rank 1 § g l + 'it'h7 2 § e 1 ! +- .

1 § g 1 + �h7 2 § d1 ! +- (the immediate 1
§ d 1 is also good) 2 . . . 'it'g7 3 �d7 § a7+ 4 'it'e6
§ a6+ 5 § d6 § a8 6 § d8 +- .
9-1 Now let us see what happens if Black is on
move.
l ... §a8+ 2 �d7 §a7+ 3 �d6 §.a6+ 4
�c7 (4 'it'c5 § e6) 4 ... §a7+ with a draw.
Let us shift all the pieces except for the black
rook a single file to left. Then the side checks do
not help anymore because the rook is not remote
enough from the white pawn : l . . . § a8+ 2 �c7
§ a7+ 3 'it'c8 § a8+ 4 �b7 +- .
In chess literature, this situation is usually
Hence we can conclude :
referred to as "The L ucena Position, " even
I ) If the pawn is on the 7th rank, multiple
though the Spaniard Lucena did not examine it in
winning methods exist. The most important ones
his book published in 1 497. The first mention of
are building a bridge for protection from checks
an analogous position was in the book by Sa1vio
along files and a rook maneuver for protection
( 1 634), which referred to Scipione of Genoa.
from side checks along ranks.
If White is on move he wins: 2) When the king of the weaker side is cut
1 ft g l + �h7 2 ft g4!
off from the pawn, the only defensive technique
2 'it'f7 is premature in view of 2 . . . § f2+ 3
consists in side checks.
'it'e6 § e 2 + 4 'it'f6 § f2 + , and the king has only 3) A rook pursuit of the enemy king can only
one way to take refuge from rook checks : by re­ be successful when the rook and the pawn are
turning to e8. The rook move prepares an inter­ separated at minimum by 3 lines. As we shall
ference at e4. This method is called building a see later, this rule does not only pertain to side
bridge, or simply bridging. checks.
2 ... ftd2 (2 . . . § a8+ 3 'it'f7) 3 �f7 §f2+ 4 4) A central or a bishop pawn divides the
�e6 §e2+ 5 �f6 ftf2+ chessboard into two unequal parts : one is "long,"
l f 5 . . . § e 1 , then 6 § g5 D. 7 § e 5 . another is "short." The correct positioning of
6 �e5 §e2+ 7 §e4 +- forces for the weaker side is to keep the king on
It is worth mentioning that White has other the short side, and the rook - on the long side.
winning options :

1 46
 Rook Endgames

The Pawn on the 6th Rank

Sax - Tseshkovsky First let us examine the situation when the

Rovinj/Zagreb 1 975 king of the weaker side is placed in front of the
pawn.

9-2
9-4

W? w

1 Elh3+?
He should not move the rook away from the Black 's rook must remain passive, staying
d-file where it was protecting the king from side on the 8th rank. White wins easily by bringing
checks. An easy win was 1 f7! E!. c8 ( l . . .�g7 2 his rook to h7 .
E!.g3+; l . . J� e 1 + 2 �f6 E!. f1 + 3 �e7 E!. e 1 + 4 �f8 1 Elb7 .§cS 2 .§ g7+1 (2 E!. h7 E!.c6) 2 . . �f8
.

E!. a 1 5 E!.h3+ �g6 6 �g8 +- ) 2 �e7 E!. c7+ 3 3 .§h7 �gS 4 f7+
E!.d7 +- . It is worth mentioning that Black can hold
1 �g6 2 .§g3+
•.•
the game when he is on move and his rook stands
Black resigned; as he failed to recognize that on a7 : l . . . E!. g7+! 2 �f5 (2 fg stalemate) 2 . . . E!.g2.
the position had become drawn: 2 . . . �h7 3 f7 E!. c8! Also, White cannot win in the case when his
(rather than 3 . . . E!. c6+? 4 �d7 +- ) 4 �e7 (4 E!. d3 king is placed on the other side of the pawn, at
�g7) 4 . . . E!. c7+ 5 �e8 E!. c8+ 6 �d7 E!. a8=. e6 : 1 E!. b7 �f8 (there are other possibilities as
well) 2 �f5 E!. a 1 ! = .
Exercises

9-5

9-3

911
W?

1 Elb7 .§cS 2 .§ g7+ �hS! (2 . . . �f8?? is er­

roneous in view of 3 �h7 +- ) 3 .§h7+ �gS =
Conclusion: passive defense holds against
a knight pawn but loses against a bishop pawn
or a central pawn.

When the stronger side has two knight

pawns, then passive defense does not help.

1 47
 Dvoretsky's Endgame Manual

3 �d8
3 <i!id6+ is useless: 3 . . . ®f6 (3 . . . <i!if8) 4 §.f7+
®g6=.
9- 6 The waiting attempt 3 §. b7 can be met either
$ with 3 . . . ®g6 4 <i!id6 ®f6 5 e7 ®f7= or with 3 . . . §.a1
4 ®d7 :8 a8 5 e7 <i!if7= (but not 3 . . . ®g8?? 4 ®f6
§ £8+ 5 §. f7 +- ) .
w In case of 3 §. d6!?, 3 . . . §. a 1 ? is bad because
after 4 ®e8 ®f6 the pawn steps ahead with a
check. 3 . . . §.b8? loses to 4 :8 d8! §.b7+ 5 <i!id6 §.b6+
6 ®d7. The only correct reply is 3 . . . ®g6!.
1 �b61 �f8 2 gS (2 g7?? �f6+!) 2 �a8 3
3 §.a7+ 4 �d6 §.a6+ (4 . . . <i!if6?? 5 §. f8 +
•••

g7 �cs 4 �f6 +- �:::. s ms+.

•••

<i!ig7 6 e7) S �e5 §.aS+! 6 �dS �as ( 6 . . . :8 a7??

7 :8 d7 + ; 6 . . . :8 a 1 !?) 7 §.d7+ (7 e7 ®f7 8 :8 d8
Now we come to positions with the king cut
§. a5 +) 7 . . . �g6! (rather than 7 . . . <i!if8?? 8 ®f6
off from the enemy pawn .
=

because passive defense does not help against a

central pawn).
The reason for the drawn final was the posi­
tion of the black rook: it was placed on the long
9- 7
side. Let us shift all the pieces except for the
black rook one file to the left. Now when the rook
is on the short side, Black, as one can see easily,
is lost.

Let us examine another position, not elemen­

tary but quite an important one.

When on move, White wins. The simplest

way begins with a check from g I , but 1 <i!ie8 is
9-8
also possible: we come to the position with the
pawn on the 7th rank and the rook on the d-file
delivering protection from side checks. For ex­
ample, l . . .<i!if6 2 e7 §. a8+ 3 §. d8 §. a7 4 §. d6+
®g7, and now either 5 §. d 1 ! +- or 5 §. e6! +- (but
by no means 5 ®d8?? §. a8+ 6 ®d7 <i!if7=).
With Black on move, the evaluation changes :
l ... �a7+ D 2 �d7 Only two fi les separate the black rook from
2 <i!ie8 ®f6 3 §. e 1 §. e7 + ; 2 ®d6 ®f8. the pawn, and this circumstance offers White
2 �a81
•••
winning chances. However a straightforward at­
The simplest defensive method : Black pre­ tempt 1 §. a 1 ? ( 1:::. 2 §. g l +) misses the win :
vents the position with the pawn on the 7th rank. l . . . §. b7+ 2 ®d8 §. b8+ 3 ®c7 §. b2 ( 1:::. 4 . . . <i!if8 or
2 . . . §. a6?? would have been a grave error in view 4 . . . ®f6) 4 §. f1 §.a2! 5 e7 §. a7+ with a draw, be­
of 3 ®e8+ ®f6 4 e7, and if 4 . . . ®e6, then 5 cause the rook managed to deliver long side
®f8! +- . checks in time.
However any other rook retreat along the a­ For a win, White should yield the move to
file, for example 2 . . . §. a 1 , does not give up the his opponent. As a matter of fact, 1 . . . :8 c8 loses
draw because after 3 <i!ie8+ (the only correct re­ to 2 §. a l ; in case of l . . .§. b l , the white rook occu­
ply to 3 §. d6!? is 3 . . . §. a8!) 3 . . . <i!if6! 4 e7 ®e6! 5 pies the important square a8; l . . . <i!ig8 2 <i!if6
<i!if8 Black has 5 . . . §.fl + ! . Here he manages to hold §.f8+3 :8f7 is also bad. Only l . . .®g6 remains for
only because of the fact that the white rook is Black but, as we shall see, this move also wors­
misplaced at d7. ens his position .

1 48
 Rook Endgames

1 �d6+! 1 �f5??
.•.

But not 1 �d7? �f6 0 2 e7 'it>f7= . After l . . .e2! 2 :§ e 1 :§ e3! 3 'it>g4 'it>e5
1 �f6 2 �d7 0 �g7 (2 . . . :§ b 1 3 e7;
••. White would have had to resign.
2 . . . 'it>g6 3 :§ a 1 ) 3 �e7! 0 2 �g3 �e4 3 �g2!
White has achieved his goal by means of The only move. Both 3 :§b4+? 'it>d3 4 :§b3+
triangulation. �c2 and 3 :§ a 1 ? :§ g6+ are erroneous.
3 ... �g6 3 .§g6+
.••

After 3 . . . :§b1 , 4 :§a8! wins: 4 . . . :§b7+ (4 . . . :§b2 After 3 . . . :§ f6 4 :§ a 1 ! the white rook, occupy­
5 'it>e8 :§h2 6 :§ a7+ �f6 7 e7 :§ h8+ 8 �d7) 5 ing the long side, assures an easy draw.
'it>d6 :§ b6+ (5 . . . �f6 6 :§ f8+ 'it>g7 7 e7) 6 �d7 4 �fl �f3 5 .§b3??
:§ b7+ 7 'it>c6 :§ e7 8 �d6 :§ b7 9 e7. And again the position is lost (a passive
4 .§a1! .§b7+ 5 �dB defense against a central pawn). Necessary was
5 'it>d6 is also good. 5 :§b2! E:: a6 6 d f2+! (we saw this stalemate when
5 .§b8+
••• discussing diagram 9-4).
After 5 . . . 'it>f6, White 's winning method is 5 .§a6 6 .§b1 .§h6 7 � g 1 .§ g 6+ White
•••

instructive : 6 e7! :§ b8+ (6 . . . :§ xe7 7 :§ fl +) 7 �c7 resigned.

:§ e8 8 �d6! :§ b8 9 :§ fl + �g7 10 'it>c7 :§ a8 1 1
:§a1 ! +- . One of the most famous "comedy of errors"
6 �c7 .§b2 occurred in the following endgame.

Capablanca - :vlenchik
Hastings 1 929
9-9

9-11
W?

B
7 .§e1!
This is the point! With the king at g7, Black
could have played 7 . . . �f8, while now the pawn
cannot be stopped.
1 .§a6?? ( l . . . :§ b8=; l . . .f:: b 1 =) 2 .§d7??
•••

7 .§c2+ 8 �d7 .§d2+ 9 �e8 .§a2 10

••.
Capablanca "amnesties" his lady rival. 2
e7 + - . 'it>f8+ wins.
2 .§a8 3 .§e7 .§a6??
•••

Trauicf)medie§ B lack repeats the same error.

4 �f8+1 �g6 5 f7 .§aS+ (5 . . . �f6 6 'it>g8!)
Uhlmann - Gulko
6 .§e8 .§a7 7 .§e6+ �h7 8 �e8??
Niksic 1 978
A single step away from reaching the goal,
White misses again. Both 8 :§ e 1 and 8 :§ f6 won.
8 .§a8+ 9 �e7 .§a7+??
•••

9 'it>g7! led to a draw.

9-10
. . .

10 �f6 Black resigned.

B?

1 49
 Dvoretsky's Endgame Manual

Alburt - Dlugy Exercises

USA eh, Los Angeles 1 99 1

9-13
9-12

912
B? B?

1 ... <cf}g5?
Black could have had an easy draw after
l . . . .§ a4!, occupying the long side with his rook.
2 E{a6! E{b4 9-14
The game was adj ourned in this position.
Grandmaster Dlugy, assisted during home analy­
sis by two experienced colleagues, Wolff and
lvanov, failed to understand the essence of the 9/3
position, and his first move after the resumption B?/Play
of play was a decisive error. What is even more
striking is that Dlugy had the classic work by
Levenfish and Smyslov on rook endings at his
disposal. In that book, naturally, the position at The Pawn on the 5th Rank
diagram 9-8 is examined. Black had to avoid that
Philidor, 1 777
position but, after a short while, it arose on the
board anyway.
3 <ct'e6
lf3 �d6 then 3 . . . �f5!= (3 . . . �g6?? 4. �c5+ ). 9-1 5
3 <cf}g6??
•••

After 3 . . . Etb7 we have the above mentioned

basic position but shifted one line down, and
this circumstance could enable Black to hold. Both
3 . . . Et b8 and 3 . . . Etbl were playable, too.
4 <cf}e7+ <cf}g7 5 f!a7! +-
Black had obviously expected only 5 e6?
Etb7+ 6 �d8 .§ b8+ 7 �c7 .§ b l = . This is the so-cal led "Philidor position."
s ... Etb8 The famous French chessplayer was the first to
5 . . . Et b6!? 6 e6 .§ b8 were more persistent; demonstrate, as early as the 1 8th century, the
White had then to employ the triangular maneu­ correct method of defense.
ver: 7 �d6+! �f6 8 �d7 0 �g7 9 �e7! 0 . l. . . E{b6! (preventing a penetration of the
6 e6 0 <cf}g6 7 E{a1 f!b7+ 8 <cf}d6 Etb6+ 9 wh ite king to the 6th rank) 2 e6 f!b1 =

<cf}d7 f!b7+ 10 <cf}c6 E{b8 1 1 <ct'c7 E{h8 I f the pawn stood at e5 the white king would
(l l . . .Etb2 12 .§ e l l) 1 2 e7 Black resigned. have had a refuge from vertical checks. But, as
soon as the pawn has stepped forward, the ref­
uge does not exist anymore.
I f White is to move in the initial position,
then, as Phil idor thought, 1 <cf}f6 wins, and his
explanation was l . . . Et fl + 2 �e6 �f8 3 Et aS+ �g7

1 50
 Rook Endgames

4 'tle7 l='1 b l 5 e6 (we know this position already : 3 § a8+ l='1 f8 4 § xf8+ 'it'xf8 5 'it'h7 +- ) 3 .§aS+
see diagram 9-8) 5 . . . l='1 b7+ 6 'it'd6 l='1 b6+ 7 'it'd7 \tle7 4 .§gS! (White prepares 5 'it'h7 ! ; the black
l='1 b7+ 8 'tlc6 +- . rook will be unable to disturb the king from the
Later on, the second defensive method in side) 4 . . .§g2 5 \tlh7! \tlf7 6 g6+ \tle7 7 .§aS
.

the Philidor position was discovered: an attack .§h2+ s \tlgS .§g2 9 g7 +-

from the rear that helps Black to hold as wel l . If But this position is also drawn. B lack's rook
the rook fails to occupy the 6th rank "a la comes in time for a passive defense along the 8th
Philidor, " it must be placed in the rear of the rank: l . . . § a l ! 2 § b6 § a8=.
white pawn.
l . . . .§ell 2 \tle6 \tlfS! 3 .§aS+ \tlg7
Now we can evaluate the position of the
black rook. It prevents both 4 'tle7 and 4 'it'd7 . Lobron - Knaak
Plus, Black can meet 4 'tld6 with 4 . . . 'tlf7 !, and Baden Baden 1 992
White must retrace his steps : 5 l='1 a7+ 'it'e8 6 'it'e6
'it'f8! etc. If he tries 4 l='1 e8, preparing 5 'it'd7, the
black rook occupies the long side: 4 . . . l='1 a l ! = .
The move 2 . . . 'it'f8! is undoubtedly correct 9-18
(the king goes to the short side, leaving the long
side for the rook), but 2 . . . 'tld8?! 3 l='1 a8+ 'tlc7 does
not lose either.
B?

9-1 6
1 \tlxe3? 2 \tlc2 .§x g3 3 .§eS .§g2+
...

Draw, according to the second defensive

method in the Philidor position.
w To avoid the theoretical draw, Black should
have played 1 . . .'it>d3!. The white king is placed at
the long side, and one cannot see how White
can survive, for example 2 § d5 + <tlxe3 3 'tlc2
4 l='1 e8 (4 'tlf6 'tld7!) 4 . l='1 h l ! (rather than
. . 'tle2!? (3 . . . § xg3 4 § e5!? 'it>f4 5 § e8 § d3! -+ , or 5
4 l='1 e2? 5 'it'f7 l='1h2 6 l='1 g8! l='1 h7+ 7 l='1g7 l='1 h8 8
. . . §d5 'it>f3 6 §e5 e3 7 'it>d3 'it>f2 -+ is also play­
'it'e7 'tlc6 9 e6 'tlc7 1 0 l='1 g l +- ) 5 § g8 § e l ! 6 able) 4 §d2+ ®f3 5 §d7, and now either 5 . . . §f8!?
§g2 'it'd8 = . 6 'it>dl 'it'f2! 7 § d2+ 'it>fl -+ or 5 . . . § xg3 6 § e7 e3
Obviously, such a defense with the king on 7 <it'd3 'it>f2 -+ followed with § f3-f8.
the long side would have been impossible if the
short side were even shorter (in case of an f- or g­ Dreev - Beliavsky
pawn). USSR eh, Odessa 1 989

9-1 7 9-19

B w

In this position, the attack from the rear does White is in a precarious situation: l . . .<it'd3
not work anymore: l ... .§gl? 2 .§a6 \tlfS (2 . . . §fl is threatened. Dreev tries his last chance.

151
 Dvoretsky 's Endgame Manual

1 h4! .§ x h4?? 2 .§ x g 5 �c3 3 .§d5! = White played 1 8 more moves before he

.§hl + 4 �e2 .§h2+ 5 �dl �d3 6 �cl .§hl + agreed to the peaceful outcome of the game; its
7 �b2 .§el S .§dS .§e4 9 �cl �e2 10 �c2 result was vitally important for both rivals.
Draw.
Black should have given a rook check and Exercises
moved his pawn to g4. Later on, he could either
trade kingside pawns, under more favorable cir­
cumstances than has actually happened, or move
his king to the g-pawn. The eventual conse­ 9-21
quences of l . . .�gl +! were :
2 ®d2 �g2+ 3 �e l g4 4 �fl (4 h5 �c3 5
h6 �h2 6 � h8 d3 7 � c8+ �d4 8 � d8+ 'itle3 9 9/4
� e8+ �f3 -+ ) 4 . . . �h2 5 � xg4 �c3 6 �gl �c2! B?
7 �g8 (7 h5 d3 -+ ) 7 ... d3 8 � c8+ 'it'b2 -+ ;
2 �e2 d3+ 3 �d2 (3 �f2 d2) 3 . . . �g2+ 4
�dl g4 5 h5 �h2 6 � c8+ (6 � xg4+ �c3 -+ )
6 . . . �d4 etc.

Larsen - Tal
Bled cmsf(9) 1 965
9-22

9-20 9/5
W?/Play

The Umbrella
The queenside pawns will inevitably be
Let us refresh our memory about the meth­
traded, and the Philidor position will probably
ods we have already seen, of sheltering the king
occur thereafter.
from rook checks.
l .§a7+ �c8?
1 ) The king approaches the rook - an effec­
The black king goes the wrong way : he
tive method when the rook is not too far away
should have tried for the short side. After 1 . . . �e8!
from the king and the pawn .
2 �e6 �f8 3 � a8+ 'itlg7 4 ®xe5 b3 5 ab � xb3,
2) "Bridge" - the rook gives protection to
the draw is obvious.
the king.
2 � xe5 b3? (as Muller indicates, after
3 ) "Refuge" - the king hides himself behind
2 . . . � h2 the position is still drawn) 3 ab .§ xb3 4
his own pawn.
�d6 .§d3+ 5 �e6?
Larsen misses his chance to punish his op­
It is a good time to show one more method.
ponent for a grave positional error and allows
Sometimes an enemy pawn can serve as a sort of
him to employ the second defensive method in
umbrella that protects the king from checks, as in
the Philidor position. The winning continuation
the next diagram.
was 5 �e7! �h3 6 �a4 ( t::. �c4+ ; �d4) 6 . . . �h7+
7 �e8 � h8+ 8 �f7 +- .
5 .§h3 6 .§aS+ (6 � a4 'itld8!) 6 . . . �c7 7
•••

.§fS .§e3! S e5 .§el 9 .§eS (9 ®f6 �d7!)

9 ... .§hl! 10 .§aS .§ell

1 52
 Rook Endgames

Velicka - Polak cause it is too close to the pawn. However when

Czech eh tt 1 995 the pawn has not crossed the middle line, such a
rook position is justified, because the rook and
the pawn are separated by no less than three
rows, and therefore pursuing the king by the rook
9-23 gives chances for a draw.

A. C heron, 1923

B?

9-25

l ... f4! 2 gf E!b2+ 3 �fl �f3 -+

White 's own f4-pawn prevents him from
saving himse1f with a check on f5 . w
4 E!al E!h2 5 �gl E!g2+ 6 �hl E!gS 7
�h2 e2 8 f5 �f2 9 �h3 E!g51 White resigned.

Traaticumedies
This is a typical case of an easy draw due to
a frontal attack.
A. Zaitsev - Hiibner
1 <it'b4 ( !:o. 2 c5 +- ) l . .. E!bS+ 2 <it'a5 E!c8!
Biisum 1 969 2 . . . §a8+? is erroneous in view o f 3 �b6 +- .
3 <it'b5 §b8+ 4 <it'a6 §c8 5 §d4 <it'e5 6
E!h4 <it'd6 =

In the initial position, the rook is placed best

at c8 where it prevents a pawn advance. How­
9-24
ever Black holds with a rook at h8, too. He meets
1 c5 with either l . . .�e7 2 �c4 :2: d8= or l . . . § h4
(cutting the king off the p aw n ) 2 c6 �e7 3 c7
.§ h8=. Horizontally cutting the king offfrom
B
the pawn is a useful defensive method.
Another important tip : in this sort ofposi­
tion, the black king should stay on the 5th or
6th rank. If he doesn 't Black usually loses.
The game continued l E!bl?? 2 �h5 E!gl
•..

(otherwise 3 �g6) 3 g5 fg 4 f5! �f8 5 f6 Black Kochiev - Smyslov

resigned. L'vov zt 1 978
A draw could have been achieved with
l . . . .§ b4 2 f5 .§ b 1 ! 3 �h5 .§g1 ! . The waiting tactic
with l . . . .§a5 was quite good, too : after 2 g5 (2 f5
.§ a 1 !) 2 . . .fg+ 3 fg Black could defend the posi­
9-26
tion either in the Philidor method (3 . . . .§ a6) or pas­
sively (3 . . . .§ a8).

w
The Pawn Hasn 't Crossed the Mid-line

In this section, we shall learn one more de­

fensive method, the one that is called "thefron­
tal attack." Both 1 �e4 and 1 .§ h 1 might lead to a draw.
If, say, the white pawn stands on b5, it makes However White carelessly moved the king away
no sense for Black to keep his rook on b8 be- from a safe place.

1 53
 Dvoretsky 's Endgame Manual

1 <;t>e2?? ®b5 2 .§.bl + ®a4 3 .§cl <;t>b4 4 1 <;t>c4 .§.c8+ 2 ®b5 .§.d8 3 ®c5 .§.c8+ 4
.§.bl + ®a3! 5 .§cl .§.d5! <;t>b6! .§.d8
First of all, Smyslov has optimally activated White has placed his king at its most active
his king (an ideal place for the king is 2 squares position. Now it i s time to protect the pawn with
away from the pawn diagonally), and now he pro­ the rook. U n l i ke the K o c h i e v v s . S m y s l o v
tects the pawn with the rook. Were the white endgame, he cannot do it horizontally. However
king at e3, he could attack the rook immediately, the rook can be placed behind the pawn here,
while now White cannot do it in time. because the black king fai l s to help to his rook in
6 <it>e3 <;t>b2 7 .§.c4 (7 �e4 §d4+) 7... <;t>b3 time.
White res i g n e d . The pawn cro s s e s the 5 .§.dl! <;t>f6 6 ®c7 .§.d5 7 <;t>c6 .§.a5
middle of the board and, with the white king on 7 . . . § d8 8 d5 is also hopeless.
the long side, the position is lost. 8 .§et!
It i s important to cut off the king from the
pawn again. Now B lack loses in view of the un­
lucky distribution of his pieces : the king stands
at the long side while the rook is at the short one.
9-2 7
For examp l e , 8 . . . § a6+ 9 �b5 § d6 10 �c5
§ d8(a6) 1 1 d5 etc .

A. C heron, 1 926*

9-29

With Black on move, this is a draw: 1 . . . �d5!

2 § c4!? �d6! 3 �a4 �d5 ! .
White, i f on move, wins. W?
l .§.c6
1 § c5 1:::.. 2 §h5, 3 b5 is no less strong.
l ... <;t>d5 2 .§.a6 .§.b7 (or 2 . . . �d4) 3 <;t>a4
<;t>c4 4 .§.c6+ ®d5 5 b5 +- The method that was applied in the previ­
Conclusion : cutting off the king of the ous example does not work here . After 1 �d4
weaker side along a rank can often be more § d8+ 2 �c5 § e8 3 �d5 § d8+ 4 �c6 § e8 5
effective than the same procedure along a file. § e 1 ? �g6 6 �d7 § a8 (or 6 . . . § e 5 7 �d6 §a5)
Black's rook occupies the long side with an obvi­
Now let us discuss situations with the b lack ous draw.
king being cut off from the pawn by more than The winning idea is to create checkmate
one file. threats to the black king that is pressed to the
edge of the board. H i s current position on h5 is
A. C heron, 1923 optimally suited for defense. Therefore White,
util izing zugzwang, must drive it away from h5 .
1 .§.g2! <it>h4! 2 .§.g7
2 e5? § xe5+ 3 �f4 is premature in view of
9-28 3 . . . �h3! = . 2 § g6 �h5 3 § d6? �g5 4 �d4 § a8=
also brings nothing.
2 <;t>h5 3 .§. gl! O <;t>h6
•.•

Now 3 . . . �h4 loses to 4 e5! § xe5+ 5 �f4.

w The idea of cutting off the white king along the
rank also does not help: 3 . . . § a8 4 e5 § a4 5 e6
�h6 (5 . . . § a6 6 �f4 § xe6 7 �f5 +- ) 6 e7 § a8 7
�f4 § e8 8 �f5 § xe7 9 �f6 +- .

1 54
 Rook Endgames

4 lit>d4 E!dS+ 5 'itlc5 E!eS 6 'itld5 E!dS+ 7 are naturally more significant. Therefore, for ex­
'it>e6! E!e8+ 8 'it>f6! +- . ample, i f the pawn stands on the 3rd rank the
king should be cut off by three files for a win
It is important to remember that in case ofa (with only two files it is a draw if, of course, Black's
knight pawn, cutting off the king by twofiles is king and rook are placed "in accordance with the
not sufficientfor a win. rules") .
It deserves to be mentioned that a frontal
A. Cheron, 1923
attack is particularly effective against a rook
pawn . For example, with a pawn on a4 even cut­
ting the king off by three files is not sufficient for
a win.
9-30
Traui�()medie§

We have seen a tragicomedy in a game by

w Kochiev, where his grave error had fatal conse­
quences. A draw would have maintained excel­
lent chances of his quali fying for the Interzonal
tournament. After losing, he failed to qualify, and
1 E!d4
the whole career of this young talented grand­
After 1 �a4 f:l. a8+ 2 �b5 .§.b8+ 3 �a5 .§.a8+
master fell steeply down thereafter.
4 �b6 .§. b8+, the king can avoid checks only by
Many have erred in similar situations, even
returning to b3 . The edge of the board is too
the greats of this world.
close, and there is no comfortable square two
steps away from the pawn diagonal ly. Tal - I. Zaitsev
1 . . . 'it>e5! 2 'it>c3 USSR eh tt, Riga 1 968
l f 2 .§. d7 , 2 . . .'�e6! 3 f:l. a7 �d6 4 <;t>a4 �c6=
follows.
2 ... E!h8
Another method of defense deserves atten­ 9-31
tion, too : 2 . . . .§. c8+ 3 .§. c4 (3 <;t>d3 .§. b8) 3 . . . .§.b8 4
.§. c6 <;t>d5 5 .§. a6 (a similar position with the king
on b3 would have been winning) 5 . . . f:l. c8+ 6 <;t>b3
.§. c6! 7 .§. a7 .§. b6= ( .6. 8 . . . <;t>c6). W?
3 E!d7
3 b5 .§. b8! 4 .§. h4 �d6! 5 �b4 <;t>c7= .
3 . . . 'it>e6! =
It should be mentioned that 3 . . . f:l. b8? loses White should play <;t>d3 and .§.b1, but which
to 4 �c4 �e6 5 f:l. a7 (5 .§. d4 +- ) 5 . . . �d6 6 b5 order of moves is correct?
.§.c8+ 7 �b4 .§.c7 8 b6. The continuation 3 . . . f:l.c8+? A theoretical draw can be achieved after 1
occurred in Dolmatov vs. Sorm (Lugano 1 986); .§. b 1 ! g5 2 �d3 f:l. e 5 3 �d4 .§. e8 4 .§.g1 �g6 5
White won after 4 �b3 <;t>e6 5 .§. d4 <;t>e5 6 .§. c4 �d3 (see the previous diagram) .
.§. b8 7 .§. c6 �d5 8 .§. a6. However 1 'it>d3?? !!el l happened, the
game was adj ourned here, and White resigned
Unti l now, we have only considered posi­ without further play. In order to bring the rook to
tions with the pawn on 4th rank. The cases of a the 1 st rank, White must attack the black rook
less advanced pawn are much more complicated, with his king, but we know that a king i s placed
and they occur much less often, therefore we badly on the 2nd rank. Here are the eventual con­
shall not investigate them. I wish only to men­ sequences if the game were continued :
tion that the distance between the pawn and the 2 'it>d2 E!e6 3 E! b l g5 4 E!gl (4 �d3 g4 5
hostile rook i s longer when the pawn stands on .§.b5 g3 6 <;t>d2 g2 7 .§. b 1 .§. g6 8 .§.g1 <;t>h5 9 <;t>e3
the 2nd or 3 rd rank, and the defending resources �h4 -+ ) 4 'it>h5 1 ( .6. 5 . . . g4)
•••

1 55
 Dvoretsky 's Endgame Manual

It is useful to improve the king's position 8 :;!e4+ 9 \t/d3 g4 10 :;!fl +

•••

by bringing him from h6 to g6. A premature 1 0 1"\ g2 .§ f4 1 1 <;t>e2 <;t>g5 1 2 1"\ f2 ( 1 2 1"\ g 1
4 . . . 1"\ e5? misses the win : 5 <;t>d3 �h5 6 <;t>d4 1"\e2 <;t>h4) 1 2 . . . g 3 ! leads to the same result.
7 �d3 .§h2 8 �e3= . 10 :;!f4 11 \t}e2
•••

5 f! h l + \t/g6 6 :;!gl ( 6 1"\ e 1 1"\ xe 1 7 <;t>xe1 1 1 1"\ g 1 <;t>g5 1 2 �e2 <;t>h4.

�h5 8 <;t>fl <;t>h4! -+ ) l l . . .g 3! 1 2 :;! x f4+ \t/ x f4 13 \tiel \t/e3!
14 \t/fl \t/f3 0 -+ .

Exercises
9-32
$

B 9-33

9/6
6 :;!e5! 7 \t/d3 \t/f5! 8 \t/d4
•••
W?
After 8 1"\ fl + �g4 9 <;t>d4 (9 1"\g1 + <;t>f3 1 0
<;t>d4 1"\ a5 -+ ) both 9 . . . 1"\ a5 1 0 <;t>e3 1"\ a3+ 1 1 <;t>e2
1"\ a2 + 1 2 <;t>e3 <;t>h3 -+ and 9 . . . 1"\e2 10 <;t>d3 1"\h2
( 1 0 ... 1"\ g2) are strong.

A Rook and a Rook's Pawn vs. a Rook

The King is in Front of Its Own Pawn

9-35

9-34

White wins, because the black king fails to

reach c7 in time.
A draw is inevitable. The only possible at­
1 f!h2 \t/e7 2 :;!h8 \t/d6
tempt to free the king from custody is the trans­
fer of the rook to b8, but then Black's king will If 2 . . . �d7, then 3 .§ b8 1"\ a 1 4 <;t>b7 .§ b 1 + 5
�a6 1"\ a 1 + 6 <;t>b6 1"\b1 + 7 �c5 . With Black's
stand in for the black rook on guard.
king on d6, the square c5 is not avai lable for es­
1 f!h2 \t/d7 2 f!h8 \t/c7 3 f!b8 :;!cl (or
caping, so White must find another itinerary.
3 1"\ h 1 4 1"\ b7+ <;t>c8 5 1"\ b 2 .§ c l ) 4 f!b2 E!c3,
. . .

and White cannot progress.

3 .§.b8 f!al 4 \t/b7 E!bl + 5 \t/c8 §cl + 6
\t/d8 f!hl 7 E!b6+ \t/c5
Let us move the black king and the white Both 8 1"\ e6? 1"\ a 1 and 8 .§ a6? 1"\ h8+ 9 �d7
1"\ h7+ 10 �e8 1"\ h8+ 1 1 �f7 1"\ a8 are useless
rook one file away, as in the next diagram .
now.
8 E!c6+! \t/b5 (8 . . . �d5 9 § a6 §h8+ 10 �c7
1"\ h7+ 1 1 �b6) 9 :;!c8 :;!h8+ 10 \t/c7 E!h7+ 1 1
\t/b8+ - .

1 56
 Rook Endgames

A standard defensive formation: Black's rook

is behind the hostile pawn, while the king is placed
Vladimirov - Rashkovsky on g7 or h7. White 's rook is riveted to the pawn
USSR chsf, Chelyabinsk 1 975 and cannot leave a8. I f 1 �b6 , then 1 J;tb1+ .
•••

T h e white king cannot escape from vertical

checks . Black's rook dri ve s the king away and
returns to a 1 .
9-36 Other defensive systems occur much less
fr e quent ly . The blac k king can hide in the
"shadow" of his opponent, or (with the black
rook on the 7th rank) in the "shadow" ofhis own
W? rook. We just mention th ese ideas but do not
study them here. Sometimes they are sufficient
for a draw, and sometimes not. For example, if we
move the black king from g7 to c3, the move 1
To achieve a draw, White should simply force f(c8! w ins A drawn position is one with the white
.

Black's king to the h-file: 1 f( g8+! �f5 2 f(f8+ king on c7 and the black king on c5 .
�g4 3 f( g8+ 'it'h3 4 f( g5 h4 5 f( g8=.
2 f(e1 = is also good.
And 1 f( c l ! t:. Back to the last diagram, let's add a white
1 §.c3? h4 2 J;te3? pawn on h5 . For the outcome, there will be no
The decisive error! It was still not too late to change : B lack simply ignores its existence. The
return to the correct plan by playing 2 f(c6+! �g5 same is valid for a g5-pawn and even for 2 or 3
3 f( c8 h3 (or 3 . . . f(h7 4 f( g8+) 4 f(h8 (4 f( g8+) white pawns on the g-file.
4 . . . �g4 5 f( g8+ 'it'f3 6 f(f8+ �g2 7 f(g8+ �h1 8 However an f5-pawn wins .
f(g6 h2 9 f( g8=.
2 J;th7! 3 J;te1 (3 f(h3 �g5 4 �e2 �g4 is
•••

also bad) 3 h3 4 J;th1 �g5 5 �e3 �h4 6

•••

�f2 J;tf7+ 7 � g1 (7 �e2 f( a7 , planning 9-38

8 . . . f(a2+ and 9 . . . h2) 7 §.a7 White resigned.
•••

IfS f(h2, then 8 . . . f(a1 + (but, of course, not

8 . . . 'it'g3?? 9 f( g2+!) 9 'it'f2 �g4 0 is decisive .
A similar position (like diagram 9-36, but with w
the white rook on d4 plus reversed w ing s and
colors) occurred in Dvoiris vs. Kovalev, USSR
ch( 1 ) , Simferopo l 1 9 8 8 . Curiously enough,
1 f6+ �f7 (l . . .�xf6 2 f( f8+; l . . .'it>h7 2 t7)
Kovalev lost precisely in the same way as
2 §.h8, and B lack loses his rook.
Vladimirov: 1 f(d3? h4 2 f(e3?.

Trastlwmedle§
The Rook is in Front of the Pawn and
the Pawn is on the 7th Rank Khaunin - Fridman
Leningrad 1 962

9-3 7 9-39

w B

1 57
 Dvoretsky 's Endgame Manual

l ... hg 2 hg? g4+! 3 fg. A draw is inevitable, fends the pawn with his king, a series of checks
as White has only g-pawns extra (no matter follows, and then the rook returns to f6. For ex­
whether one or two). ample, 3 �d5 .§b6 4 �c5 .§f6! (the best place
The winning continuation was 2 �xg3! �h7 for the rook ! ) 5 �b5 .§f5+!, etc .
3 h4! gh+ 4 �h3 �g7 5 f4, and the f-pawn goes
ahead with a decisive effect . Now let us move the white king to f4.

The Rook is in Front of the Pawn and P. Romanovsky, 1950

the Pawn is on the 6th Rank

J. Vancura, 1 924*
9-41

9-40

B
l . . .� fl +? 2 ®e5 � f6 is bad here on account
of 3 � g8 + ! . However Black has no other defen­
The main difference between this position sive plan than the rook transfer to the 6th rank.
and those discussed above, is the fact that here Therefore l . . . .§cl!
White 's king has a refuge from vertical checks : lf 2 ®e5, then 2 . . � c6= follows, achieving
.

the a7-square. The king hides there in order to the Vancura position. White may use the oppor­
free his rook from the job of protecting the pawn. tunity for removing his rook from the corner.
The black king, in contrast, fails to reach 2 .§b8 .§al 3 .§b6 (3 � b7+ �f6 4 a7 ®e6
the queenside : 1 . . . �f7? 2 �e4 (2 a 7? ®g7 would is weaker.)
have been premature) 2 .. .rtle7 3 a7! �d7(f7) 4 When the rook stood on a8, the black king
�h8 +- . was riveted to the kingside; but now the time
2 . . . � a 5 (instead of 2 . . . �e7) is also hope­ comes for a march to the pawn . But this should
less: 3 �d4 �g7 4 �c4 �f7 5 ®b4 � a 1 6 ®b5 be done carefully: the premature 3 . . . �f7? 4 ®e5
�b1 + 7 �c6 � a 1 8 �b7 � b 1 + 9 ®a7 ®e7 1 0 �e7 5 � b7+ ®d8 6 a7 loses for Black.
� b8 � c l !? 1 1 ®b7 (rather than 1 1 � b6? �d7) 3 . . . .§a51 4 �e4 �f7! 5 �d4
1 l . . .� b 1 + 12 ®a8 � a 1 13 a7 +- . Black's king If 5 � h6, then 5 . . . �g7 ! , but not 5 . . . ®e7? 6
fails to reach c7 in time (see diagram 9-3 5). a7 ®d7 7 � h8 ! .
Because of this analysis, the diagrammed 5 ... �e7 6 �c4 �d7 7 � b4 .§ a t , and the
position had been considered winning for a long draw is clear.
time. However a saving plan was finally discov­ It is worth mentioning that l . . .§b1? (instead
ered. This plan is based on the fact that the a6- of l . . .� c l ) would lose: 2 � a7+! 'it>f6 (2 . . . 'it>g6 3
pawn gives the king a refuge from vertical checks, � b7 � a 1 4 � b6+ ®f7 5 'it>e5 + - ) 3 �e4 �b6 4
but cannot hide him from side checks. Therefore � h7! �g6 5 a7 � a6 6 �b7 +- .
Black should bring his rook to f6 . However the first moves might have been
l ... .§fl +I 2 �e4 .§f6! . This is the so-called transposed: l . . . � a5!? 2 �e4 �c5! (2 . . . �b5!) 3
" Vancura position . " Black follows the same �a7+ ®g6 4 �b7 (4 �d4 � c6=) 4 . . . �a5=.
"pawn in the crosshairs" method found in end­ In many lines, the kings compete in a race to
ings with bishops of opposite colors. The rook the queenside. I f the white king stood closer to
attacks the pawn in order to prevent the enemy 's the pawn, then the black one would eventually
rook from leaving a8 . What can White do? If a6- arrive too late . This means that Black cannot de­
a7, Black always has � a6 (his king will obviously lay the rook transfer to the 6th rank; this plan
never leave the g7- and h7-squares). If White de- should be executed as soon as possible.

158
 Rook Endgames

S a7 E!.a4 6 E!.c7+ �g6 7 � x b2 �f6 8

�b3 E!.a1 9 �b4 �e6 10 �bS ( 1 0 'it>c5)
Vyzhmanavin - Lerner 10 �d6 11 E!.c6+ (l l fl.h7) ll ...�dS 12 E!,a6
•••

USSR eh, L'vov 1 984 E!.b1 + 13 �aS �cS 14 E!.c6+! �xc6 1S aS�+,
and White won.

Brodsky - Khmelnitsky
9-42 Kherson 1 989

B 9-44

In this drawn position, Black uses his last W?

available trap, and unexpectedly succeeds.
1 ... E!.a3!? 2 �d2??
Correct was, of course, 2 \t>b2! .§. f3 3 'it>c2
\t>d5 4 \t>d2= . 1 .§.a4+! \t>f5 2 .§. c4 .§.a2+ 3 'it>g3 .§.b2 4 .§. a4
2 h 2 ! 3 �e2 E!.a1! White resigned.
••• would have led to a draw.
1 E!.g6+? �f4 ( l . . .\t>f5?? 2 El.g3= is the
Ivanchuk - Lautier Vancura position) 2 E!.a6
Horgen 1 995 White has n o 2 El. g 3 o n account of
2 . . . .§.gl +!. 2 .§. f6+ can be met by 2 . . . 'it>e4!, while 2
.§. c6 - by 2 . . . .§. a2 + !.
2 ... �e4 3 E!.a4+ �e3?
9-43
An absurd move ! If the king is going to
move ahead, then why not to d3? But 3 . . . 'it>d5 4
.§. f4 El. a2+ 5 ®g3 .§.b2 would have been a much
simpler win.
B

9-45
1 E!,b4??
•••
$
The Vancura position could be achieved
through the elementary l . . J�( f6! 2 El. a8+ (after 2
\t>xb2 Black's rook would become "desperado" W?
because his king is stalemated) 2 . . . \t>xh7 3 'it>xb2
\t>g7.
2 E!.c7 (2 .§. b7 is also good) 2 E!,a4 3•••

E!.c8+?? Now we have (with reversed wings and col­

White makes his adversary a nice present: a ors) the Romanovsky position ( 1 950). Its solu­
vitally important tempo. 3 a7?? .§. xa7= was obvi­ tion is 4 .§.h4!! 'it>d2 (4 . . . .§. a2+ 5 'it>gl ! .§. f2 6 .§. a4
ously erroneous, but the line 3 El. c6! \t>xh7 4 'it>xb2 El.a2 7 .§. h4=) 5 El.h3! 'it>c2 6 .§. f3! a2 7 .§.£2 +! (7
\t>g7 5 'it>b3 .§. a l 6 \t>b4 'it>f7 7 'it>c5(b5) 'it>e7 8 El. a3?? 'it>b2) 7 . . . 'it>d3 8 .§. f3+ 'it>e4 9 El.a3=.
\t>b6 'it>d7 9 \t>b7 led to a win. 4 E!.g4? (a decisive error) 4 E!,a2+! S �h3
•••

3 ... �xh7 4 E!.c6 E!.b4?? After 5 'it>gl 'it>f3! 6 .§. a4 (6 .§. c4 .§. e2) Black has
A present in return ! After 4 . . . 'it>g7! Black's a pleasant choice between 6 . . . El. a l + 7 ®h2 \t>e2
king could have come to the queenside in time: 5 and 6 . . . 'it>g3 7 'it>fl .§. a l + 8 ®e2 a2 D. 9 . . .§. h l .
.

\t>xb2 'it>f7 6 'it>b3 .§.al 7 'it>b4 'it>e7 8 'it>b5 \t>d7=.

1 59
 Dvoretsky 's Endgame Manual

5 Etf21 6 Eta4 a2 7 Eta3+ 'if/d4 8 'if/g3

••• White could move his rook to a more
Eib2 9 'if/f4 Elf2+ advantageous position : 1 l"!. c5 ! (or 1 l"!. b5+)
White resigned. The aim of the last check l. . . �b6 2 l"!. c3 l"!. f2+ (2 ... l"!. f4 3 l"!.h3) 3 �b3 (3
was probably to improve the rook position after l"!. c2!? D. 4 l"!. h2) 3 . . . l"!.h2 4 l"!. c4 and 5 a4 with an
10 �g3, bringing it to c2 first, and to push the easy win. However Suetin does not suspect any
king thereafter. However the immediate 9 . . . �c4 danger of a draw.
10 �e4 �b4 1 1 El. aS �c3 -+ was sufficient for a 1 a4?? Sif4! = 2 'if/b3 'if/a61 (the threat was
win. 3 l"!.b5+ followed by 4 h5 +- ) 3 a5 Ete4 4 'if/c3
Sif4 5 'if/d3 f!g4 6 'if/e3 f!c4 7 'if/f3 Etc3+
a- and h-Pawns The rook is placed best on the c-file. 7 . . . l"!. b4?
loses to 8 l"!. h8 �a7 9 h5 l"!.b5 10 h6 +- .
In the Vancura position, let us add a white 8 'if/e4 Etc4+ 9 'if/d5 Etg4 10 'if/e6 Etc4 11
pawn on the h-file. It is easy to see that the evalu­ labS 'if/b7 12 f!h7+ 'if/a6 Draw.
ation remains unchanged. The defensive method
is precisely the same as before : the rook main­ SzabO - Thkmakov
tains the pawn in the crosshairs from the side Buenos Aires 1 970
and does not release the hosti le rook from the
corner.

9-48

9-46

White may simply wait and maintain the a­

pawn in the crosshairs. For example, 1 l"!. b5 �d6
l . . . Ete4+! 2 'if/d3 f!e5 3 'if/c4 f!f5! 2 l"!.f5 l"!.a1 3 �h2! a4 4 l"!.f4! a3 5 l"!.f3! �c5 (5 . . . a2
The best policy is to pay no attention to the 6 l"!. a3) 6 l"!. b3 �c4 7 l"!.f3 �b4 8 l"!.f4+! etc.
h-pawn at all . 3 .. . l'h h5? loses to 4 a6 l"!. h6 5 'lt>b5 Szab6, an experienced grandmaster, did not
l"!. h 5 + 6 'lt>b6 l"!. h6+ 7 �b7 . know this defensive system.
4 a6 ( 4 �b4 l"!. f4+) 4 §f6! 5 'if/b5 f!f5+
•••
1 'if/g2?1 'if/d6 2 'if/f2?! Eta2+ 3 'if/et?
6 'if/c6 f!f6+ 7 'if/d5 f!b6 etc. After 3 �gl ! �c6 4 l"!. f5! the position was
still drawn.
T..-al!ic()medle§
•••3 Etal +! 4 'if/e2
There is no salvation anymore: 4 �f2 a4 5
Suetin - F. Portisch
l"!. xh4 a3 6 l"!.h3 (6 l"!. a4 �c5 -+ ) 6 . . . a2 7 l"!. a3
Belgrade 1 977
l"!. h 1 , or 4 �d2 l"!. h 1 ! 5 l"!. xa5 h3 6 l"!.h5 h2 D.
7 . . . l"!.a1 -+ .
•••4 a4 5 f!h6+ 'if/e5 6 Eih5+ 'if/f6 7 'if/f2
a3 8 'if/g2 §cl 9 Eta5 Etc3 White resigned.
9-4 7

Twenty-three years later, precisely the same

position occurred in Emms-Riemersma, Gausdal
1 993 . And again, White did not know the theory
W?
of this ending.

1 60
 Rook Endgames

1 �h6+?! �d5 2 �h5+ �c4 3 � xh4+? (3 �aS �b3 1 1 �e2 a2 1 2 �d2 � c4 -+ ) 5 . . . a4! 6
�f5! �a1 4 �h2 a4 5 �f4+ with a draw) 3 �b3.•• �h3+ �b2 7 � h4 � a2! 8 'it'h1 (8 �g4 �b3+ 9
4 �h5 'it'h3 �c2 -+ ) 8 . . . � a 1 + 9 ®g2 a3 1 0 �h3 a2 -+ .
5 �g2?
An error in return. A draw was possible
through 5 �g5! a4 6 �g3+ �b4 (6 . . . �c2 7 �g2+
9-49 ®d3 8 �g4 a3 9 �g3+ �c2 1 0 �h2=) 7 �g4+
'it'b5 8 Elg5+ �c6 9 � g6+! (the king should be
driven as far away as possible; premature is 9
�g4? a3 1 0 'it'h2 �b1 1 1 � a4 �b3 1 2 �g2 �b5
B? 1 3 � a8 �b4 1 4 ®f2 �c3 1 5 �e2 �b2 -+ , or 1 0
�g2 �b5 ! 1 1 � g 3 �b4 1 2 �g4+ �b3 1 3 �g3+
'it'b2 -+ ) 9 . . . �d5 10 �g5+ �e6 11 �g2 �b1
(l l . . . a3 12 :S g3 f:::.. � f3=) 1 2 �a5 �b4 1 3 �f2
'it'd6 1 4 ®e2 = .
4... �a1? s . . . a4 -+ 6 �f2 ( 6 �h3+ �b2 7 � h4
White 's rook is misplaced, and it is impor­ a3 -+ ) 6 ... a3 7 �b5+ �a2 8 �e2 �b1 9 �d5
tant to keep it on the h-file. Emms demonstrated (9 � a 5 �b2 10 ;; b 5 + �a1 1 1 � a5 a2 -+ )
that this could have been achieved by the subtle 9 ... �b2 10 �d2+ �b3 11 �d3+ �a4 12 �d4+
move 4 . . . � a3!, for example: 5 �h2 (5 �g5 �b4+ �b4 13 �d8 a2 14 �d3 �b3 White resigned.
6 �g2 � c3 7 �g8 a4 8 �b8+ �c4 9 �f2 a3 1 0

A Rook and Two Pawns vs. a Roo k

As Tarrasch once said, "all rook endings switches to the second defensive method in the
are drawn. " These endings are rife with drawish Philidor position ( i f 3 �c6 then 3 . . . :S xd4 4 § b8+
tendencies, and even as large a material advan­ �e7 = ) .
tage as two extra pawns is often not sufficient for 3 d6 �g1!
a victory. The rook prepares itse lf for giving r e a r
checks because the white king has no refuge at
Doubled Pawns d6 anymore .
4 �c6 �cl+ 5 �d5 �h1 6 �b8+ �d7 7
If the king ofthe weaker side stands in front �b7+ �d8 8 d7 �h5+ (8 . . . �e7) 9 �c6 �h6+
of the pawns, a draw can usually be easily 10 �c5 �h5+ 11 d5 �h6! =

achieved (except for those cases when the rook And again, B lack returns to the defensive
must stay on the back rank in view of mate method suggested by Philidor.
threats). The applicable ideas here are fami liar to
us from the Philidor position (diagram 9- 1 5 ). Connected Pawns

Two extra connected pawns can be most

easily exploited if the king supports them. H ow­
9-50 ever, sometimes they can advance for queening
even when a rook alone supports them, as in the
next diagram.

1 �b7 �g6 (l . . . �c8!?) 2 �b6 �g4!

Not the only move, but the safest. B lack

161
 Dvoretsky 's Endgame Manual

Szabo - Keres 6 ®f6 f! xg5!= (or 6 . . . f! f5 + ! ) .

Moscow 1 95 6 6 �f4 �b4+ 7 �e4 �b6 =

Let us look at a more compl icated but quite

useful situation that can occur in a practical game.
9-51
G Kasparian, 1 946

w
9-53

1 �g6
White is planning 2 h6, 3 g5 , 4 f! g7, 5 h7 w
etc . This simple plan cannot be prevented . Such
pawns are sometimes called "self-propelled. "
l...�e4 2 h6 �h2 3 g5 �d3 4 �g7 �c3
5 h7 ( .6. g6, f! g8) 5 . . . �b3 6 �b7+ B lack re­
signed. For the present, let us accept that White
wins if he succeeds in transferring his rook to the
The best chances for a successful defense 5th rank. This means that B lack dare play neither
exist when the king blocks the pawns. This is l . . .®f8 2 f! f7 + .6. . . . f! f5 nor l . . . !'! d3 2 f! c7 f!h3
perhaps the most important drawn position : (2 . . . f!d5 3 h6) 3 f!c5. Therefore his rook must
stay on g3 and h3 . But what can Black do when
J. Kling, B. Ho rwitz, 1 851
the white king comes to the kingside?
It turns out to be difficult for White. If his
king comes to g2 when the black rook is on h3 ,
9-52 then l . . . !'! a3 is playable, because 2 f! b7 f! a 5 ! 3
h6 f! g5 + loses a pawn ; the same happens after 2
®h2 !'! b3 3 f! a7 f! b 5 ! .
If White plays 2 'it'f2 the rook goes back to
w h3 . By the way here, as in all similar positions, 3
h6 f!h5 4 'it'f3 f! g5 5 f! g7+ ®h8 6 ®f4 f! f5(g4)+ !
leads t o nowhere.
H owever if, with the white king on f2 and
1 �d4 �b6 2 �d8! �b4+ 3 �e5 �b7! the black rook on h 3, Black is on move he comes
The most precise: B l ack protects the 7th to be in zugzwang. His rook must leave its com­
rank and threatens to take the g-pawn (he cannot fortable position behind the pawns, and then the
of course do it immediately: 3 .. .'it>xg5?? 4 h7) . white rook has the opportunity to leave h7.
Erroneous is 3 . . . f!g4? 4 f!g8+ 'it'h7 5 'it'f5! +­ We have come to the conclusion that f2 and
(rather than 5 f!g7+?! ®h8 6 'it'f5? f!f4+! - a "des­ h3 are the squares of the reciprocal zugzwang.
perado" rook). The Encyclopaedia ofChess End­ Obviously enough, another pair of such squares
ings claims that 3 . . . f!b5+ 4 f! d5 f!b7 5 'it'e6 also is e2 and g3 . Furthermore, when the white king
loses for Black. But I do not see how White can stands on any dark square of the 2nd rank the
make any progress after 5 . . . f! a7 6 f! d7 f! a6+ 7 black rook must be on h3 while, when the king
f!d6 (7 'it'e7 'it'xg5 8 h7 !'!aS 9 'it'f7 <it>h6=) 7 . . . f!a7. stands on a light square, the rook must be on g3 ! .
5 . . . f!b6+ 6 'it'e7 f!b7+ 7 f! d7 f!b5 (7 . . . f!b8) 8 h7 1 �a2!!
f!b8! = is also good. A paradoxical move that contradicts the
4 �g8+ �h7 5 �e8 �g6 standard approach ("first we move our king to
Black returns to the initial position of thi s the kingside, and only think thereafter"). It turns
ending. But he can now force a draw with 5 . . . §b5+ out that one should be thinking immediately be-

1 62
 Rook Endgames

cause any other initial move misses the win. (rather than 9 . . . .!::! h4? 1 0 �g2 0 +- ) 1 0 �g2
After 1 �b2? §h3 2 �c2 § g3 3 �d2 §h3 4 .!::! h4 0 , and the rook cannot be forced away from
�e2 § g3 S �f2 §h3 White is in zugzwang: 6 the h-file.
�g2 §a3 7 § b7 § a S = . 7 ... Etdl + 8 <it'c6 Stet+ 9 <it'd6 l=tdl + 10
l f l § b7?, then l . . . § gS 2 § h7 §g2! 3 �bl l=td5 §al ll <it'e7 Eta6
§h2 4 �cl §g2 S �dl §h2 6 � e l § g2 7 �fl White 's task is less difficult in case of
§h2 (the same zugzwang position, only by a rank l l . . . .!::! e l + 1 2 �d8! �g8 ( 1 2 . . . .!::! al 1 3 l::! d 7+ �g8
lower) 8 �gl §a 2 9 § b7 § a S = . 14 �e7 +- ; 1 2 . . . �h6 1 3 .!::! d 7! �xh5 1 4 g7 .!::! gl
I n case of 1 �bl ? , l . . . § g2? is erroneous: 2 lS �e8 �h6 1 6 �f8 +- ) 1 3 h6! (13 .!::! f5!? .!::! e6
�c l § h 2 3 �dl §g2 4 � e l § h 2 S �fl , and 14 �d7 .!::! a6 lS .!::! cS +- ) 13 . . . .!::! e6 C l 3 . . . l::! hl 14
Black is in zugzwang: S . . . §a2 6 § b7 +- . The cor­ �e7!; 13 . . . .!::! g1 14 .!::! d6) 14 h7+! (but not 14 .!::! gS?
rect method is l . . . § b3 + ! . The rook gives checks .!::! a6 l S �e7 §b6 16 h7+ �g7 17 .!::! hS .!::! b7+ 18
unti l the king steps on the 2nd rank, and then �e6 .!::! b6+ with a draw) 14 ... �h8 1 5 .!::! gS �g7
goes to a corresponding square . For example 2 16 .!::! h S +- .
�c2 §g3!, or 2 �cl §c3+! 3 �d2 (3 �dl § d3 + !) 12 §d7
3 . . . § h3 ! 4 �e2 § g3 5 �f2 § h3 0 = .
l .. .§h3 2 <it'b2 §g3 3 <it'c2 §h3 4 <it'd2
§g3 5 <it'e2 §h3 6 <it'f2 0 §a3 7 §d7 §h3 8
§d5 <it'g7 9 <it'g2 §h4 10 <it'g3 +- 9-54
It remains for us to prove that White wins if $
he succeeds in bringing his rook to the 5th rank.
This fact is not quite obvious because Black
blocks the pawns with his king. However his B
blockade is less efficient than in the Kling and
Horwitz position.
1 <it'a2!! §d3 2 §b7! (but, of course, not 2
�b2? § h3! 0 ) 2 §h3
= ••• 12 ... Etb6!?
2 . . . § g3 can be met by 3 �b2 .!::! g5 4 .!::! h 7 12 . . . .!::! aS 13 �e6+ �g8 14 h6 +- is quite
.!::! g 3 S �c2, and it is Black who turns out to be in bad. After 1 2 . . . .!::! c6, Kasparian gives 13 �d8+
zugzwang again. This is the simplest way, but �g8 14 .!::! e7 �f8 ( 1 4 . . . .!::! d6+ l S �c7 .!::! a6 1 6
another, more universal way also exists : 3 .!::! b3!? �d7 1:::. 1 7 .!::! e6) l S �d7 .!::! a 6 1 6 .!::! e6 .!::! a7+ 17
.!::! g S 4 .!::! h3 �g7 (otherwise the white king goes �d6 .!::! a6+ 1 8 �eS .!::! aS+ 1 9 �f6 +- .
ahead) 5 h6+ �g8 6 g7! (rather than 6 h7 +? �h8 13 <it'd8+ <it'g8
7 .!::! h6 .!::! gl and White 's king will not have a ref­ 13 . . . �f8 is met with 14 �c7! .!::! a6 1 5 h6 O S
uge from rook checks from the rear) 5 . . . �h7 6 .!::! d6) 1 S . . . .!::! xg6 1 6 .!::! d8+ and 1 7 h7 +- .
�b3, White activates his king and gradually wins 14 <it'c7! Eta6 15 Etd6 1:::. h6 +-
(a similar position was analyzed by Kling and In Theory of Rook Endings by Levenfish
Horwitz as long ago as in 1 85 1 ). and Smyslov, in the very end of this line, another
3 §b5 <it'g7 (3 ... l::! g3 4 .!::! b3) 4 §g5! road to the win is suggeste d : 1 4 .!::! e7 �f8
Now Black has neither 4 . . . �h6 S g7! nor (14 . . . .!::! d6+ lS �c7 .!::! a6 16 �d7 .!::! b6 17 .!::! e6 +- )
4 . . . .!::! c3 S h6+! �xh6 6 g7. But this position is l S .§ f7+ �g8 1 6 �e7 § a6 1 7 .!::! f6 .!::! a7+ 1 8 �e6
winning for White even without this move (when .!::! a6+ 1 9 �fS .!::! aS+ 20 'it'g4 �g7 2 1 .§f7+. But
the black king stands on h6). this recommendation is erroneous : instead of
4 ... §h4 18 . . . .!::! a6+? Black plays 18 . . . �g7!, because after
In case of 4 . . . �g8 S �b2 .!::! e3 White plays 1 9 .§ f7+? .!::! xf7 20 gf �f8 0 he holds a pawn
6 �c2! .!::! a3 7 �d2 .!::! b 3 8 �e2 .!::! a3 9 �f2 .!::! b3 ending despite being two pawns down.
10 .!::! d5 +- . A hasty 6 h6? .!::! h3 7 h7+ �g7, on the This complicated analysis can hardly be
contrary, leads to a theoretical draw. (and certainly should not be) remembered in all
5 <it'b3 l=th1 6 <it'c4 Stet+ 7 lit'd5 its details. To know that the rook transfer to the
The king must go ahead. Nothing can be 5th rank wins is quite enough, yet the proof of
achieved by 7 �d3 .!::! h l 8 �e3 .!::! h3+ 9 �f2 .!::! hl this fact turns out to be rather complicated.

1 63
 Dvoretsky 's Endgame Manual

Tr-aui�()medle� f- and h-Pawns

Glek - Leitao Endings with these pawns are mostly drawn.

Wijk aan Zee 1 999 Their theory is rather complicated and that is why
we will explain only the basic ideas here. The
following example from practical play shows how
one should defend these positions.
9-55
Gligoric - Smyslov
Moscow 1 947

9-58

1 ... .§c1+?? 2 <it'g2 .§c2+ 3 <it'g1 .§h2 4

.§b6
The Kasparian position - Black has no win. w
4... <it'f7 5 .§h6 <it'e7 6 .§g6 <it'f8 7 .§f6+!
<it'e8 8 .§e6+! <it'd7 9 .§h6! h3 10 .§g6 Draw.
Black could have won by means of 1 . . .h3! 2
�g4+ �g6 3 � h4 (3 � a4 g2) 3 . . . h2+ 4 <it>g2 �h6 The black rook is excellently placed on the
5 � g4+ <it>f6 6 �f4+ (6 <it>h1 � g6 +- ) 6 . . . <it>g5 7 5th rank: it prevents an advance of the hostile
� fl <it>g4 8 � a 1 h1�+ 9 � xh1 � xh1 1 0 <it>xh1 king. lf 1 f5, then l . . . � b l , threatening a series of
<M3 (Glek). checks from the rear.
1 .§g6+ <it'f7!
Exercises l . . .®h7 was not losing, but B lack would
have had more problems than in the game.
2 .§g5 .§b1!
A typical retreat for this sort of situation:
the rook maintains opportunities for checks from
9-56 various directions, both from the side and rear.
3 .§c5
In case of 3 h6, 3 . . . � g l +? is erroneous : 4
917 ®f5 � h 1 5 � g7+ ®f8 6 ®g6 � g 1 + 7 ®h7! � f1 8
W? � a7! (8 � g4 ®t7) 8 . . . � xf4 9 ®g6 � g4+ 1 0 <it>f6!
� f4+ ( 1 0 . . . <it>g8 1 1 � g7+!) 1 1 ®g5 � £1 1 2 � a8+
<it>t7 13 h7 +- .
The waiting move 3 . . . � a l ! helps, for ex­
ample: 4 �h5 (4 ®f5 � a S + ; 4 h7 �gl + 5 <it>f5
� h l ) 4 . . . <it>g8 5 f5 ®h7.
The following attempt is interesting: 4 �g7+
<itl£6 5 �c7!? <it>g6 (Black can also play 5 ... �g1 + 6
9-5 7 <it>f3 �hl) 6 h7, (see next diagram).

9/8
B?

1 64
 Rook Endgames

E!h3 21 �d5 E! d3+ 22 �c6 E! c3+ 23 �b7 E!h3

24 �c8 +- .
Instead of 6 . . . E!g1 +? Black must play 6 . . . E!h1 !
9-59 immediately. If White tries the waiting move 7
$ E!b7, Black can wait too: 7 . . . E!h2, with no fear of8
E!b5 �g7! 9 E! g5 + �h8 ! . Another good line is
7 . . . E!g1 + 8 �f3 E! h 1 9 �e4 E! e 1 +, because when
B the white rook stands on b7 the king 's route
around it is too long: 10 �d5 E!d1 + 1 1 �c6 E!c1 +!
1 2 �b6 E!h 1 ! with a draw.
If 7 �f3 , Black can play either 7 . . . �f5! or
I once investigated the position that arises 7 . . . �f6!? 8 �e4 E! e 1 + 9 �d5 E!d1 + 10 �c6 E!h1 ! .
after 6 .. J::! g1 +? 7 �f3 E! h 1 8 �e4 E!h5 together When the black king stands o n f6, White has no
with grandmaster Gulko. The continuation 9 f5 +? important move 1 1 E! e7, while after 1 1 E! d7 (or
�f6 10 E! c6+ �g7 1 1 �e5 �xh7 (l l . . .E! h 1 !? is 1 1 �b7) 1 l . . .�f5 the black king abolishes the
even simpler) 12 E! c7+ �h6! 1 3 �f6 E!h1 14 E! c2 f4-pawn and returns to g6 in time.
E!h3!= does not promise White any success (see 3 .. \t>f6 4 .§.c6+ \t>g7!
.

the ending below, Polugaevsky - Ree) . The main danger for B l ac k is for his king to
White should cede the right to move to his be forced to the back ran k . This could have hap­
adversary, with the idea afforcing the black rook pened after 4 . /�f7? 5 �g5 2 g l + 6 �f5 E! h l 7
.

away from h5, and then advancing his king, when E!c7+.
his rook stands precisely on c7. All this can be 5 \t>g5 .§.g1 +! 6 ®f5 .§.a1 7 .§.c7+ (7 E!g6+
achieved in the following way: 9 E! d7(e7) �f6 �f7) 7 ... \t>h6 8 .§.e7 .§.b1 9 . .§.e8 \t>g7 10 .§.e5
1 0 E! a7 (10 E! d8? E! xh7 1 1 E! d6+ �e7=) 1 0 . . . �g6 .§.a1 11 .§.d5 .§.f1
( 1 0 . . . E! h 1 ? 1 1 E! a8! E! xh7 1 2 E! a6+) 1 1 E! c7 ! ! 0 Not a bad move, but holding the rook in the
E! h 1 (l l . . .�f6? 1 2 E! c8! +- ) 1 2 �e5 E! e l + 1 3 corner was quite enough .
�d6 E! d 1 + ( 1 3 . . . E! h 1 1 4 E! e7! A 1 5 �d7) 1 4 12 .§.d4 .§.a1 13 .§.d6 .§.a5+ 14 <;t>g4 .§.a1
�c6 E! h 1 ( 1 4 . . . E! c 1 + 1 5 �b7 o r 1 5 �d7) 1 5 E! e7! 14 . . . E! b5 is also playable : it leads back to
�f5 1 6 �d7 �xf4 1 7 �e8 �g5 1 8 �f8 �g6 19 the initial position.
�g8 +- . 15 .§.e6 .§.g1 + 16 \t>f5 .§.a1
Later, I discovered the possibility of a more
stubborn defense. Instead of 1 4 . . . E! h 1 , Black
should play 14 . . . E! d8!?.
In order for White to win, he need only get
9-60
his king back to the f-pawn, while keeping the
Black rook tied to the 8th rank. But how is this to
be accomplished? Black answers 1 5 �c5 with
1 5 . . . E! a8!, after which 16 �d4 is useless: 16 . . .
w
E! a4 + 1 7 �e3 ( 1 7 �e5 E! a 5 + 1 8 �e4 E! h5)
1 7 . . . E!a3+ 1 8 �e4 E! h3 (A 1 9 . . . E! h5) 1 9 �e5
E! e3+ 20 �d5 E! d3+ 2 1 �c6 E! d8!, etc.
Before bringing the king back, it's impor­
tant to bring the rook to d7 first. Then Black's
17 h6+ \t>h7! 18 .§.d6 .§.a2 19 \t>g5 .§.g2+
rook maneuver to h3 (as in the variation we just
20 \t>f6 \t>xh6! 21 \t>e7+ \t>h7 22 f5 .§.e2+ 23
examined) has no point - once again, White
.§.e6 .§.a2 24 f6 .§.a8!
brings his king forward, and now the Black rook
We have discussed this sort of position in
cannot get to d8. The most exact line is: 16 E! a7!
the section dedicated to the pawn on 6th rank.
(not 1 6 E! d7 at once : 1 6 ... E! a 5 + 1 7 �d4?! E!h 5 , The black rook is placed on the long side, so a
and White must start all over again) 1 6 ... E! b8 1 7
draw is inevitable.
E! d7 �f6 ( 1 7 . . . E! a8 1 8 �d4) 1 8 �d4 (threaten­
ing �e4-f3-g4) 18 . . . E! b4+ 19 �e3 E! b3+ 20 �e4

1 65
 Dvoretsky 's Endgame Manual

25 ctlf7 ctlh6 26 Elel Ela7+ 27 Ete7 EtaS the loss only for a little while: 3 �g6 �gl + 4 �f7
2S Eld7 ctlh7 29 Eldl Ela7+ 30 ctle6 Eta6+ 31 � a l 5 �g8+ �h7 6 � e8 � a7+ 7 �f8. White 's
Eld6 EtaS 32 Eld4 ctlgS 33 Elg4+ ctlfS Draw. next move will be 8 f7 (the h5-pawn deprives the
In this example, Black kept his king on f7 black king of the important g6-square ) .
until the danger of its being driven to the back 3 h6
rank arose. Thereafter the king went to g7 and Black resigned; he cannot prevent h6-h7.
later on - to h6, attacking the wh ite pawn. But,
strictly speaking, Kopaev 's recommendation was
to place the king in front of the more advanced
Polugaevsky - Ree
pawn.
The best position for the rook is on a I ; it is Amsterdam 1 98 1
ideal ly suited for giving checks along files as
well as ranks. However, if the pawns are not ad­
vanced too far, the rook stands quite well on the
9-62
5th rank, and sometimes on fl .

It goes without saying that not all positions

with f- and h-pawns are drawn. The most impor­
w
tant exception was already mentioned above :
Black usually loses if his king is cut off on the
back rank.
l ElaS?!
Capablanca - Kostic The simplest way i s to keep the rook on h8,
Havana m ( I ) 1 9 1 9 in order to profit by the side checks in case of
emergency.
l E{g2+ 2 ctla3 ctlc3 3 El x a4
•.•

It was still not too late to bring the rook to

9-61 the long side, for example 3 �h8!? c4 4 �h3+
�d2 5 �b4!=. However this capture does not
lose, contrary to comments by Krnic in the Chess
Informant.
w 3 c4 4 EtaS Elg7 5 ctla2??
•••

This is the decisive error! Now the black

king advances while the white rook remains
chained to the a-file. He should have followed
One does not need to keep the solution in the waiting policy: 5 � a6! � d7 6 � a8 � d l
mind because White has many winning ways to ( 6 . . . �c2 7 � h8!) 7 �a2 �c2 8 �h8 c 3 9 �h2+
choose from. with a draw.
l f6 5 ctlc2 -+ 6 ctlal c3 7 ctla2 Elb7 S Eta6
•••

Kopaev suggests 1 � b8+ �h7 ( 1 . . .�£7 2 Eld7 9 EtaS ctld2 White resigned.
h6 b. 3 h7; l . . .�g7 2 f6+! � xf6 3 h6+) 2 f6 �c5+
(2 . . . � c7 3 � e8) 3 �g4 � c4+ 4 �f5 �c5+ 5 �e6 Other Pairs of Disconnected Pawns
� c6+ 6 �e7 � c7+ 7 �f8 �h6 8 f7 +- .
As a rule, two extra pawns are sufficient for
l ... E{cl 2 Elg7+
a win. However exceptions occur now and then.
Belavenets's suggestion is also good: 2 h6
They are caused either by the stronger side hav­
�gl + 3 �f5 � fl + 4 �e6. The king is striving for
ing badly placed pieces while the defender 's
the 8th rank. If 4 . . . � e l +, then 5 �d6! (rather
pieces are active, or by inattention (when the
than 5 �d7? �f7 6 h7 � h l =) .
stronger s i d e anti c ipates a quick w i n too
2 ctlfS?!
nonchalantly). This last case is illustrated by all
•••

Loses at once, but 2 . . . �h8 could postpone

the practical examples that follow.

1 66
 Rook Endgames

Tr-aeicf)medies Kasparov - Short

London m (9) 1 993
Bernstein - Smyslov
Groningen 1 946

9-65

9-63

B
The rook that blocks a passed pawn can­
not, as a rule, leave its post unpunished. There­
fore it would have been wise to play for zugzwang:
l b2?? (both l . . .'it? e 5 and l . . . �e4 won
• • • 1 .§a2!? ®f5 2 e4+ ®e5 3 ®e3, and now 3 . . . .§ a8
elementarily) 2 .§ xb21 <it'g4 (3 . . . § c5 4 ®d3) 4 aS § h8 5 § a4 (or 5 § f2).
The planned 2 . . . .§h2+ turned out to be inef­ Another winning method was 1 �e2 ( D.
fective because of the stalemate after 3 �f3 .§ xb2. �d3-c4-b4) l . . .�e4 2 § fl ! D. .§f4+ (2 §h1 §h5!
3 <it'fl 3 § f1 !).
Draw. The Philidor position has arisen. 1 e4?? <it'e6??
B oth opponents are hypnotized by the
A similar story happened in the following above-mentioned rule. However this was a proper
endgame. moment for neglecting it (there are no absolute
rules in chess ! ) by playing l . . . § c 5 ! . Black could
Gufeld - Bronstein then regain a pawn and block the a-pawn again
Kislovodsk 1 968 in time, for example 2 a5 (2 §a3 §c4 3 a5 § xe4 4
a6 §£4+ D. § f8=) 2 . . . §c3+ 3 ®g4 (3 �e2 ®xe4
4 a6 §c8=) 3 . . . ®xe4 4 a6 §c8 5 a7, and here the
9-64 most precise defense is 5 . . . .§ g8+! 6 ®h5 § a8,
although 5 . . . § a8 6 § a5 �d4 7 �f5 �c4 8 ®e6
�b4 9 § a 1 ®c5! is also sufficient for a draw
(rather than 9 . . .'it?b5? 10 �d6 �b6 1 1 §bl +!),
W? e.g. 1 0 �d7 �b6 11 §b1 + ®c5! 12 §b7 § h8=.
2 <it'e3 <it'd6 3 <ifjld4 <it'd7 4 <ifjlc4 <it'c6 S
<it'b4 .§eS 6 .§cl+ <ifjlb6 7 .§c4 Black resigned.

Larsen - Torre
With 1 ®t7! , White maintains his two extra
Leningrad izt 1 973
pawns : l . . .<it>g4 ( l . . .®e3 2 .§ h2) 2 .§ d4+ <it>f5 3
c4 .§ c7+ (3 . . . ®e5 4 .§d5+ ®e4 5 .§ g5) 4 �e8!
®e5 5 .§ g4 �f5 6 ®d8! with an easy win.
1 <it'g7? <it'g4 9-66
Now 2 .§ d4+ �h5 3 c4 can be met by
3 . . . .§ xc4! 4 .§ xc4 - stalemate.
2 .§h2 <it'g31 3 .§hl .§ xc2 4 hS .§c7+ S
<it'f6 .§c6+ 6 <it'f7 .§c7+ 7 <ifjle6 .§c6+ 8 <it'dS w
.§h6 9 <it'e4 <ifjlg2 10 .§h4 <it'g3 11 .§hl <it'g2
Draw.

1 67
 Dvoretsky 's Endgame Manual

A natural method of exploiting two extra The rook cuts the hostile king off from both
pawm is a transition to theoretically winning pawns. White cannot strengthen his position.
positions with one extra pawn. 7 <l;'f4 §.e2 8 E!d5 §.c2
This method could be applied here: 1 �g5! 8 . . . ®xf6? is erroneous: 9 .§. c5 .§. e8 1 0 .§.f5+!
.§. xc5 2 'it'g6 b. .§. h8+. White has a simple win �g6 ( 1 0 ... �e7 11 .§.e5+ �f7 1 2 .§. xe8 �xe8 13
because the black king is on the long side. 'it'e5 +- ) 11 .§. e5! .§. c8 1 2 .§. e6+ 'itlf7 1 3 �e5 +- .
1 §.c7?! <l;'d8 2 §.c6 <l;'d7 3 §.d6+ <l;'e7 4 9 §.d6 §.e2 10 f7+ <l;'xf7 11 <ct'f5 <l;'e7 12
f6+?? §.d7+ <l;'e8 13 <l;'f6 §.e1 14 E!d5 §.cl 15 §.d6
After 4 .§. e6+! �f7 5 c6 .§.fl 6 'it'g5 .§.f2 7 E!fl + 16 <l;'e6 §.et+ 17 <ct'd5 §.dl + 18 <l;'c5
.§. d6 Black would have had no alternative to a §.xd6 19 <l;'xd6 <l;'d8 Draw.
resignation. 4 .§.d5 +- was also good.
4 ... <ct'f7 5 c6 <l;'g61 6 <ct'f3 §.ell = Exercises

9-6 7 9-68

9/9
w B?

A Far Advanced Passed Pawn

Transition to a Rook vs. Pawns I n case o f the straightforward l . . . hg?

Endgame ( l . . .'it'd3? 2 .§. f2! or 2 .§.g2! has the same conse­
quences) 2 �xg3 'it'd3 3 .§.a2 c3 4 h4 c2 5 .§. xc2
It often happens that a passed pawn is so 'it'xc2 White, of course, cannot play 6 h5?? .§. d4!,
strong that the opponent must inevitably give a but 6 �g4? �d3 7 h5 'it'e4 8 �g5 �e5 9 'it'g6
rook away for it. I n such cases, one should know 'it'e6 1 0 h6 .§. g 1 + also loses. It is shouldering
well and take into account the methods we have that helps here : 6 �f4! �d3 7 h5 .§. h 1 8 �g5
l e a r n e d from study i n g rook versus pawns �e4 9 h6 'it'e5 10 �g6 �e6 1 1 'it'g7! (rather
endgames. than 1 1 h7? .§.g1 + 1 2 �h6 �f7 1 3 h8.£\+ �f6 1 4
Black's actions in the following endgame ®h7 .§.g2 0 -+ ) 1 1 . . .'it'e7 (l l . . . .§. g 1 + 1 2 'it'f8)
were based on two typical methods : shouldering 1 2 h7 .§.g1 + 1 3 �h8! = .
and cutting off the king. D e l i b erat i n g over h i s n e x t m o v e ,
Tseshkovsky recognized White 's defensive plan
Yusupov - Tseshkovsky
and found how to prevent its realization.
Moscow tt 1 98 1 l ... §.fl +I! 2 <l;'g4 hg
Now, after 3 �xg3 'it'd3 4 .§.a2 c3 5 h4 c2 6
.§. xc2 �xc2, the white king cannot go to f4, and
White loses.
9-69 3 §.d2+ <ct'e3 4 §.g2
4 .§. c2 .§.f4+! 5 �xg3 .§. d4 6 h4 'itld3 changes
nothing.
4 §.f4+1
•••

B? White could hold after 4 . . . c3? 5 .§. xg3+ 'it'd4

6 .§. g2 �d3 7 h4 (compared with the line 3 'it'xg3
�d3 he would have an extra tempo). Alas, Black

1 68
 Rook Endgames

wins easily by means of cutting the king off along

the 4th rank.
5 � xg3 c3 6 h4 Etc4 7 Etc2 �d3 8 Et cl 9- 71
c2 9 h5 �d2 10 Ethl cl� 11 Et xcl �xcl!
White resigned.

The most important method in sharp end­ 8

ings with a far-advanced passed pawn is inter­
ference ("bui lding a bridge"). It occurs, together
with other useful techniques, in the following
example. 2 . . . .§ a 1
2 . . . g4 is very bad in view of 3 § cS + and 4
§aS (a bridge again). The same method decides
Balashov - Dvoretsky
in case of 2 . . . 2: b4+ 3 �aS §b1 4 § c4 ! .
USSR eh tt, Moscow 1 967*
After 2 . . �h4 White can apply another typi-
.

cal method : deflection of the rook, namely 3 §c4+

§ xc4 4 a8'§' . H owever after 4 . . . § f4 a theoreti­
cally drawn position ari ses (we will study this
9- 70 sort of ending later in the book) . Therefore the
interference method should be applied here, too :
3 �bS ! § a 1 4 § c4+ and 5 .§ a4 .
3 § c8!
B? 3 § cS? § xa7 is erroneous: the rook is placed
badly on the 5th rank, and even more, it stands in
the way of the white king.
3 . . . § xa7
The main threat is by no means �cS-b6-b7 Equivalent is 3 . . . g4 (or 3 . . . �g4) 4 a8'l* § xa8
- in that case the king will certainly be late when S § xa8.
coming back to fight against the black pawn . 4 �xa7 �g4
White is planning 2 �bS ! followed by the inter­ Or 4 . . . g4 S �b6 g3 6 § g8! (6 �cS? �g4!=)
ference : § c6+ and § a6. If l . . .�fS?, then again 2 6 . . . �h4 7 �cS �h3 8 �d4 g2 9 �e3 �h2 1 0
�bS ! § a 1 (2 . . . § xa7 3 § x a7 g4 4 �c4 �e4 S �f2 +- .
s �b6 �f3 6 § f8 + !
§ g7 �f3 6 �d3 g3 7 § f7 + and 8 �e2) 3 § cS + !
A familiar method: zwischenschach for gain­
�f4 4 § c4 + a n d S § a4 , or 3 . . . �f6 4 § c6+ and
ing a tempo.
5 § a6.
6 . . . �e3 7 § g8! �f4 8 �cS g4 9 �d4 �f3
Every tempo counts i n such situati ons.
1 0 �d3 g3 1 1 § f8+ �g2 1 2 �e2 +- .
B lack holds by means of driving the king away
by vertical checks. The king should be driven as Tl"al!i�()medie§
far as possible from the g-pawn.
l ... EtaU 2 �b6 Etbl+! 3 �c6 Eta1 4 �b7 Peters - Browne
Etbl + (the immediate 4 . . . �fS is also sufficient USA eh, South Bend 1 98 1
for a draw) 5 �c8 Eta1 6 �b8 �f5 =
Another method of preventing the threat of
interference, l . . .�hS?, looks less attractive : the
9- 72
king on the h-file will be unable to render shoul­
dering to his opponent. In reality, this move loses,
and its eventual consequences are quite instruc­
tive:
W?
2 �b6 ( ..6. 3 § c8)

1 69
 Dvoretsky 's Endgame Manual

Remembering the previous example, we can Exercises

easily find the correct solution; it is based upon
driving the king away by vertical checks: 1 l"l c8+!
�d3 2 l"l b8 (2 l"l d8+) 2 . . . �c2 3 l"l c8+! �d1 4
l"l b8 �cl 5 f5 (5 �g6) 5 . . . b1 � 6 fl xb l + �xb1 7 9- 74
�g6=.
l f5?
White fails to tackle a relatively simple prob­ 9/1 0
lem. The attempt to set another pawn in motion B?
also loses : 1 �g6? l"ld1 2 l"l xb2 �xb2 3 �xh5
�c3 4 \t'g6 \t'd4 5 h5 �d5 6 f5 (6 h6 \t'e6)
6 . . . \t'd6 7 h6 (7 f6 �e6 -+ ) 7 . . . \t'e7 8 h7 l"lg1 + 9
�h6 \t'f7 1 0 h84J+ �f6 1 1 \t'h7 �xf5 1 2 4Jf7
�f6 -+ .
1 ... .§d1 2 .§ xb2 (the same is 2 �g6 b1� 3
l"l xb1 l"l xbl) 2 ... �xb2
In the 1 l"l c8+! line, the same position oc­ 9- 75
curs, but with the king on b 1 : one square farther.
This tempo turns out to be decisive.
9/1 1
3 �g6 �c3 4 �xh5 �d4 5 �g6 �e5 6
W?
h5 .§g1 + 7 �f7 �xf5 8 h6 .§a1 9 h7 .§a7+ 10
�g8 �g6 White resigned.

Tarrasch - Bliimich
Breslau 1 925

9- 76
9- 73

9/1 2
W?/Play
w

Lasker 's Idea

Tarrasch resigned! He saw that his king was
cut off from his own pawn along the 4th rank, Books on chess endings contain many in­
while the attempt to advance the pawn 1 h6 would teresting and instructive rook-and-pawn endings
have been met by 1 ... .§b6 2 .§h5 a2 3 h7 .§bS with a single pawn on each side. We have al­
(and, if 4 l"la5, then 4 . . . a1 � 5 l"l xa 1 �xa1 6 �g4 ready studied some typical methods, that are char­
l"lh8 -+ ). acteristic for this material, in the previous sec­
The grandmaster had completely forgotten tion of this book. Now we shall discuss one more
the possibility of deflecting the black rook from idea. The second world champion introduced it.
the 8th rank: 4 .§b5+! .§ xb5 5 h8�+ .

1 70
 Rook Endgames

Em. Lasker, 1890 game I. Zaitsev - Dvoretsky, Moscow eh 1 973 . I

did not know the endgame study by Keres and,
having discovered the same idea, executed it in a
slightly different way: 1 §. a S + 'it'h4! 2 §. a 3! 0 .
9- 77 The game continued 2 . . . 'it'g5 (in case of 2 . . . 'it'g4
3 'it'f7, we transpose into the main line of the
Keres' study) 3 §.g3+ 'it'f4 4 §.g2 'it'f3 5 §. h2
(Keres suggests 5 §.b2 'it'e3 6 'it'd7 §. d 1 + 7 'it'c7
W? §. c l + 8 'it'b7 a1 '!!¥ 9 e8'1!¥+) 5 . . . 'it'e3 6 §.b2 0 �e4
7 §.e2+ �d3 8 �d8 'it'xe2 9 e8'1!¥+, and my op­
ponent resigned after a few more moves.
2 ... li£7g4 3 ®f7! §fl + 4 1i£7g6 §e1 5 §a4+
Were Black on move, he could hold the game li£7h3 6 1i£7f6 §fl + 7 li£7g5 §g1 + B li£7h5 §e1 9
by playing l . . .'it'a7! or l . . .l'� b2 ! . But it is White f!a3+ li£7g2 10 §xa2+
who i s on move, and he sets into motion a In the study by Lasker, this was the termi­
mechanism that gradually drives the black king nation point; but here the fight goes on.
as far away as the 2nd rank. 10 ... ®f3 1 1 §a7
1 1i£7b8! §b2+ 2 ®aS §c2 3 §h6+ ®a5 4
®b7 §b2+ 5 ®a7 §c2 6 E!h5+ ®a4 7 ®b7
§b2+ 8 ®a6 §c2 9 §h4+ ®a3 10 ®b6 §b2+
11 ®a5! §c2 12 §h3+ ®a2 13 § xh2 +- . 9- 79

A slightly more complicated version of the

same idea is demonstrated in the following ex­
ample. B

P. Keres, 1947*

ll . §e6! 12 ®g5 ®e4 13 §b7(c7)!

..

1 3 §. d7 ? i s erro n e o u s on account o f
9- 78 1 3 . . . �e5 0 =.
$ 13 ... 1i£7e5 C l 3 . . . 'it'd5 1 4 'it'f5) 14 §d7
Now it is B l ack who h a s fal l e n into
zugzwang.
w 14 ®e4 15 §d1! ®f3 16 §f1+ li£7e2 17
•..

E!f7 1i£7e3 18 ®f5 +- .

1 §a3 1i£7h4! Exercises

B lack prevents the rook transfer to the 2nd
rank: 2 §.h3+ and 3 §.h2.
2 §a5! 0
When the black king is placed on h4, 2 'it'f7? 9-80
is senseless - after 2 . . . §. f1 + 3 'it'g6 §.g1 + 4 �h6
§. e 1 White's king has traveled too far away from
the e7-pawn and cannot protect it. Therefore
White waits: he realizes that Black's king is placed 9/1 3
worse o n whichever square other than h4. B?
It is worth mentioning that the diagrammed
position occurred, with reversed colors, in the

171
 Dvoretsky 's Endgame Manual

A Rook and Two Pawns vs. a Rook and Pawn

All Pawns are on the Same Wing l"l e8+ 'it'd3 1 2 'it'f3 'it'd2 1 3 �f2 'it'd1 1 4 l"l f8
l"l d2+ 1 5 'it'fl l"lh2 1 6 l"l f5 �d2 1 7 l"l xg5 'it'e3 1 8
If all pawns are grouped on the same wing l"la5 Draw.
then a draw is the most probable outcome. Even If 9 . . . E!e2+, White should play 1 0 'it'f3
when there is a passed pawn, defense is, as a l"le3+ ( 1 0 . . . l"l h2 1 1 l"l f8+! fol lowed with l"l h8) 1 1
rule, not too difficult. 'it'g2=.
It is worth mentioning that here again, as on
Smyslov - Keres move 3 , a retreat of the king to the back rank
USSR eh, Moscow 1 949 loses.
10 <iftgl? f3 1 1 E!a4 E!g2+! ( l l . . .l"l e4? 1 2
l"l a 2 'it'f4 1 3 'it'f2 l"l e 2 + 1 4 l"l xe2 fe 1 5 h4! =) 1 2
<ifth1 ( 1 2 'it'fl l"lh2 1 3 l"l a 5 + 'it'g6 -+ )
9-81

9-82
w

B?

Black is planning l . . .f4+ and 2 . . . 'it'h4. A reli­

able method of preventing a king invasion is a
rook check from h8. This position occurred in Schmidt-Plachetka,
1 E!bS! f4+ 2 <iftg2 E!c2+ 3 ®f3! Decin 1 976, with a single unimportant difference :
3 'it'g 1 ? loses to 3 . . . 'it'h4 4 l"lb3 l"l e 2 /:::,. the white rook stood on b4.
5 . . . l"le3. 12 ... g4! 13 hg+ (13 l"\ xg4 l"\ xg4 14 hg+
3 ... E!c3+ 'it'xg4 -+ ) 13 ... <iflg5!
There is no danger for White in 3 . . . l"lh2 4 In case of 13 . . . l"l xg4? 1 4 l"la2 'it'f4 15 �h2
l"l h8+ 'it'g6 5 l"l g8+ 'it'f6 6 l"l h8 'it'g7 7 l"lh5 'it'g6 8 l"l h4+ 16 'it'g1 'it'g3 White holds the endgame
l"l h8 'it'f6 9 'it'g4! (9 l"l h5? �f5 1 0 l"l h8 l"\ xh3+!) because of a stalemate : 17 l"l g 2 + ! .
9 . . . l"lg2+ 1 0 'it'f3 l"lg3+ 11 'it'f2, and 1 1 . . .'it'g6 Now B lack threatens 1 4 . . . l"le2 fol lowed by
can be met, besides the waiting 1 2 'it'fl , even . . . 'it'h4-g3. As we know, a passive defense with
with 1 2 h4 g4 1 3 l"lg8+ (or 1 3 l"lf8 l"lf3+ 1 4 �g2). the rook on the 1 st rank does not help against an
4 ®g2 E!g3+ 5 ®h2 E!e3 6 <iftg2 f-pawn. As for checks from the rear, B l ack will
6 l"l h8+ 'it'g6 7 h4! g4 8 l"l g8+ 'it'f5 (8 . . . 'it'h5 use the g-pawn as an umbrella against them.
9 l"l h8+) 9 l"lg5+ 'it'e4 10 l"l xg4= . 14 §a1 E!e2!
6 ... <iftg6 7 E!fS! In the game P lachetka choose an erroneous
The simplest solution: White cuts the en­ continuation 14 . . . 'it'h4?, and White managed to
emy king off from the center of the board. hold the game by means of 1 5 l"l fl ! l"l g3 (if
7...E!e2+ 8 ®f3 E!h2 9 E!hS <iftg7 10 E!h5 15 . . . �g3 1 6 l"l g 1 �f2 1 7 l"l a 1 l"l xg4, a stalemate
<iftf6 11 E!hS E!h1 12 <iftg2 E!d1 13 E!fS+ <iftg7 saves White again: 18 l"la2+ 'it'g3 19 'it'g1 l"l b4
14 E!f5 E!d2+ 15 <iftf3 E!d3+ 16 <iftg2 <iftg6 Draw. 20 l"lg2+!) 1 6 g5! 'it'xg5 17 l"l a l .
Cutting the king off along the f-file is not However a step b y the king t o the opposite
obligatory (even more so because B lack can over­ direction would have led to a win: 14 . . . 'it'f4! 1 5
come it). Instead of 7 l"l f8, 7 l"l a8 'it'f5 8 l"l a5 + g 5 ( 1 5 l"l f1 l"le2 1 6 g 5 'it'g3 1 7 l"l g 1 + 'it'h3 1 8
l"l e 5 9 l"l a8 i s possible. The game Timman­ l"l fl f2 , o r 1 5 l"l g 1 l"l e2) 1 5 . . . 'it'g3 ( 1 6 . . . l"lh2+ 1 7
Radulov, Wijk aan Zee 1 974 (with reversed col­ �g1 f2+ i s threatened) 1 6 l"lg1 'it'f2 1 7 l"l a 1 l"l g4!
ors and wings) went 9 . . . l"ld5 10 l"l f8+ 'it'e4 1 1 1 8 l"la2+ 'it'g3 1 9 �g1 l"l b4 (the g-pawn is still on

1 72
 Rook Endgames

the board, so there is no stalemate possibility) go ahead when the white rook is placed on the g-
20 .§ a 1 .§ b2 2 1 g6 §g2+ 22 'it'fl §h2 -+ . file: 6 . 'it'h5 7 l::'!. g5+.
. .

15 E!g1 �h4 16 g5 �h3 (16 . . .f2) 17 E!a1 6 ... .§d3 7 .§g2 .§d6
E!h2+ 18 �g1 f2+ -+ . lf7 . . .f4, then 8 l::'!. g4 (8 § f2 § d4 9 'it'h3 'it'h5
10 l::'!. d2 ! = is also good) 8 . . . l::'!. d2+ 9 'it'g1 (9.'it'h3)
Vaiser - Dj uric 9 .. .f3 1 0 § f4 § d3 11 l::'!. g4 /:::,. 'it'f2 = .
Szirak 1 985 8 �h3 .§f6 9 .§ g5!
Draw in view of 9 .. .f4 10 'it'g2 f3 + 11 'it'f2
§ f4 1 2 § g3 'it'h5 1 3 § g5 + ! .

9-83 Trauif:;umedie§

J. Polgar - Short
Monaco bl 1 993
W?

9-84

In comparison with the previous ending, the

black pawns are less advanced. This circumstance
seems to be in White 's favor, but actually he is B
faced with severe problems. His rook cannot
reach h8, as with Smyslov's defensive method
against a king penetration via the h-file.
For example: if 1 .§ b6? then l . . .f5 + 2 'it'f4 After the waiting move 1 . . . §. b7 the position
§ aS 3 § b7 .§ a4+ 4 'it'g3 § a3+ 5 'it'g2 (5 'it'f4 is still drawn: 2 'it'e6 l::'!. b4 (2 . . . .§b1 is equivalent)
§ h3) 5 . . . 'it'h5 6 §h7+ 'it'g4 7 § h6 § a6 -+ . 3 .§ d7+ 'it'h6 4 �f5 §b5+ 5 'it'xf6 § b4! 6 g5+ (6
In Gliksman-Novak, Stary Smokovec 1 976, 'it'f5 l::'!. f4 + or 6 . . . § x g4 leads to stalemate)
the same position with reversed colors arose. The 6 ... 'it'xh5 7 §h7+ 'it'g4 8 g6 l::'!. b6+ 9 ®f7 l::'!. b7+ 10
game continued 1 h5? g5! ( l . . .gh+ leads to a 'it'g8 l::'!. b 8+ 11 'it'g7 ®g5 = (Muller).
drawn endgame with f- and h-pawns) 2 § b6 § f7 Short decided to at least prevent the king
3 .§ a6 'it'g7 4 'it'f5 .§ b7 5 h6+ ( 5 .§ aS 'it'h6! 6 from invading at e6, but the remedy proved worse
'it'xf6 §b1 7 ®f5 'it'xh5) 5 . . . 'it'xh6 6 l::'!. xf6+ ®h5 7 than the disease - his resourceful adversary
'it'e5 l::'!. b 3 8 § fl 'it'h4 9 l::'!. h l + l::'!. h 3 and Black found an elegant forced win.
won. l. .. E! e 7 ? ? 2 h 6 + ! �f7 (2 . . . 'it' x h 6 3
Vaiser discovered a new defensive method 'it'xf6 +- ) 3 g511 fg 4 E!d81 +- .
for this sort of ending, and thus a highly important
one : Hebden -Wood
1 �h3!! f5 2 .§a3! Hastings 1 994/95
The immediate 1 l::'!. a3 l::'!. b8 2 l::'!. f3 l::'!. b 6! 0 3
'it'g3 (3 l::'!. fl? f5+ 4 'it'h3 l::'!. b 3+ 5 'it'h2 l::'!. b4 -+ )
3 . . . 'it'h5 4 'it'h3 l::'!. a6!? 5 l::'!. fl ! (5 'it'g3? f5 !:::.
6 . . . l::'!. a 4 -+ ) 5 . . . l::'!. a3+ 6 'it'h2 f5 7 l::'!. g 1 = was good 9-85
enough, too.
..• 2 .§f7
2 . . . § b8 was more sensible, because here
White, if he wished, could have played 3 l::'!. a8 B
transposing to the plan we already know.
3 .§b3!? .§e7 4 .§g3! .§e8 5 .§g1 .§e3+ 6
�h2
-
It becomes clear that the black king cannot 1 �xf4??
•••

1 73
 Dvoretsky 's Endgame Manual

A terrible error! The black king will be cut cause 1 1 ®fS is met by 1 1 . . .§.aS+.
off along the f-file now, and the g7-pawn will be In the line 2 ... 'it'f8 3 h6 'it'g8 4 .§. g7+ ®f8 S
inevitably lost. .§. g6 'it'f7 6 h7 §. a4+ 7 'it'g3 .§. a3+ 8 'ifth4 §. a 1 9
The simplest way to a draw was 1 . . . 'it'g6 2 h 8 4J + 'it'f8 1 0 §. x f6 + 'it'g7 1 1 'it'gS ® x h 8
'it'g4 (2 fS+ 'it'xfS 3 .§. xg7 .§. a 1 4 .§.f7+ 'iftg6=) (1 l . . .§.aS+? 1 2 .§.fS .§. xfS+ 1 3 gf'it'xh8 14 'it'f6! +- )
2 . . . .§. a4 3 .§. d6+ 'iftf7. 1 2 'it'g6 White, according to the Encyclopaedia
Another way was l . . .g6 2 .§.f7+ <ifte4 3 'it'g4 of Chess Endings, should win. However it is a
(3 .§.f6 .§. a 1 4 'it'g3 .§.g1 + S ®h2 .§.g4; 3 fS gf 4 g6 mystery to me how he can do it after 1 2 . . . 'it'g8 1 3
f4 S g7 .§. g3) 3 . . . .§. a 1 4 .§. e7+ 'it'dS S .§.g7 'ifte4! 6 .§. b6 ( 1 3 g 5 .§. b 1 1 4 .§. xa6 §. b8=) 1 3 . . . 'it'f8 14 g5
.§. xg6 .§.g1 +, and in case of7 'it'h4?? 'it'f3! White 's aS.
king will be checkmated. Also playable is 2 . . . §.gS 3 §. xa6 \ftg7 4 §.a7+
2 §.£7+ �e5 3 �h51 +- §.a6 4 E!xg7 §a5 (4 h6+ 'it'g6!=, but not 4 . . . '\t'xh6? S .§. xf6+ .§.g6 6
5 §.e7+ �f5 6 E!f7+ �e6 7 E!fl §.aS S g6 'iftfS! +- ) 4 . . . \ftg8 s h6 .§. g6 6 h7+ 'it'h8 7 §.f7
§.hS+ 9 �g5 �e7 10 §.et+ (10 g7) lO �fS •••

11 �6 §.h6 (1 L..§.h7!? 1 2 .§.e8+!) 12 §.e2 Black

resigned.
9-8 7
C higorin - Tarrasch
Budapest 1 896

B?

9-86

Black should not cling to the f6-pawn. He

achieves a draw by means of 7 . . . §.gS! 8 §. xf6
B §.aS (8 . . . §. g7 9 §.h6 §. a7=) 9 §. h6 (9 §. f7 §.a4+
1 0 'it'gS §. xg4+) 9 . . . .§. a4+ 10 'it'gS §.aS+ 1 1 ®h4
.§.a7= or 1 1 . . .§.a1=. The other way Black can save
himself was shown by David Navara: 7 . . .fS! 8 gS
(8 gf §.g4+!) 8 . . . .§.a6=.
In the game, Black let the hostile king pen­ After 7 . . . .§.h6? 8 'iftf5! Black loses. The game
etrate into his camp; this caused a rapid loss. Malisauskas-Sandler (USSR 1 977) continued:
•••l §al?? 2 �f5 E!fl + 3 �g6 §f4 4 g5! 8 . . . .§.h4 (8. . . §.hS+ 9 'it'g6! §.gS+ 10 'ifth6 +- ) 9 §.d7
fg ( 4 . . . .§. xh4 S .§. a8+ 'it'e7 6 gf+) 5 hg E!a4 (9 .§. xf6? §. xh7!=; 9 g5? fg 10 'iftg6 §. f4=) 9 . . §.h6 .

Both S . . . 'it'g8 6 .§. a8+ .§. f8 7 .§. xf8+ '\t'xf8 8 10 §.e7 (a simpler way is 1 0 §. a7 0 §.h4 1 1 gS! fg
'it'h7 and S . . . .§. g4 6 .§. xa6 .§.g1 7 .§.a8+ 'ifte7 8 1 2 'it'g6 +- ) 1 0 . . . §.h4 1 1 gS §.h5 1 2 'it'g6??
.§. g8! are hopeless. A gross error when just a step away from a
6 §.aS+ �e7 7 �h6 a5 S g6 §.al 9 g7 win. 1 2 §. e8+! 'it'xh7 1 3 'iftxf6 was decisive.
E!hl + 10 �g6 §gl + 11 �h7 §hl + 12 �gS 12 . . . .§. xg5+ 13 'it'h6 .§.eS! 1 4 .§.f7 §. e8 1 5
§.al l3 §.a7+ �eS 14 §.a6 §.hl (14 . . . 'it'e7 1 5 'it'g6 §.d8 Draw.
'it'h7 .§.h1 + 16 .§.h6) 1 5 §.xa5 §.e1 16 E!h5 §.gl Exercises
17 §.e5+ �d7 1S �h7 Black resigned.
The rook had to watch the 5th rank. The a6-
pawn is not necessary for Black: its existence is
not essential for a draw. 9-88
After 1 . . . 'it'g8! 2 hS, the most solid defense
was suggested by Fridstein: 2 . . . .§. b5!? 3 .§. xa6
'it'g7 4 .§. a7 + 'it'g8 S h6 .§. cS 6 §. g7 + 'it'f8 !
( 6 . . . 'ifth8? 7 .§. f7 §. c6 8 'iftf5 +- ) 7 §.g6 ( 7 'it'g3
9/14
§. c l =) 7 . . . '\t'f7 8 h7 §. c8 9 §.h6 'it'g7 (9 . . . §.h8? 1 0
W?
'it'fS 'iftg7 1 1 §.h1 'it' f7 1 2 §. a 1 +- ) 1 0 §. h 1 .§. a8
and White cannot strengthen his position, be-

1 74
 Rook Endgames

Pawns on Opposite Wings V. Smyslov, 1976

A common situation is when one side has

two connected passed pawns while the adver­
sary has a far-advanced pawn on the opposite 9-90
wing. In these endgames, correct placement of
one's pieces is highly important.

N. Grigoriev, 1 936* 8

9-89 White has three finely placed connected

passed pawns, but sti ll the win is problematic.
t .§.b3+ 2 �d4 (2 �d2 l"l b4 3 f5 l"lb5= is
.•.

no better.) 2 ... .§.b4+ 3 �c3 .§.xf4 4 .§. xb2 .§.h4!

w 5 l3.b7+ �g8 6 .§.b8+ (6 h6 l"l g4) 6 �g7!•.•

But not 6 . . . �f7(h7)? 7 g6+ coming to a win­

ning Kasparian position (see diagram 9-5 3 ) .
7 h6+ �g6 8 .§.g8+ �h7 9 l3.g7+ �h8 t0
White has so-called self-propelled pawns. .§.e7 .§.g4 1 1 .§.e5 �h7 t2 �d3 �g6 =
However, a lot oftime is required for promotion, so Karsten Muller has showed that White still
Black manages to create counterplay in time. has a complicated path to victory. He suggested
t b6 ( .6. 2 l"la5+, 3 l"lb5) 2 �e4 l"l b4+ 3 �f5 l"lb5+ 4 �e6 (4 �g4 l"l b4 is
1 l"l d6 l"lb2+ 2 �e3 h4 does not bring any useless) 4 . . . l"l b6+ 5 �d5 l"lb5+ 6 �c6 l"l b4 7 f5
success, either. l"l g4 8 h6+ �h7 9 l"l xb2 l"l xg5 10 l"lf2 +- . IfBlack
t �f4 2 a5 .§.b2+ 3 �et (3 �g1 �g3)
••• temporizes : 7 . . l"l b8 8 h6+ �h7, then the most
.

3 h4 4 .§.a7 h3 5 l3.h7 h2 6 .§. xh2

••• exact way is 9 �d5! (but not 9 �c5? l"l g8! 10 g6+
6 �d1 is met by 6 . . . �g3! 7 �c1 l"lb5. �xh6 1 1 l"\ xb2 �g5 12 l"l f2 �f6=) 9 . . . l"\ b4 1 0
6 .§. xh2 7 b7
••• �e5 (and now, thanks to zugzwang, Black must
A typical situation: the rook cannot stop allow the king back into the lower half of the
the pawns, but Black nevertheless manages to board) 10 . . . l"lb5+ 1 1 �f4 l"l b4+ 1 2 �g3 l"lb5 1 3
hold by pursuing the hostile king, which is �g4 l"lb3 1 4 �h4 0 +- .
pressed to the edge of the board. I will note here, that with Black's pawn on
7 �e3 8 �fl (8 �d1 �d3 9 �cl �c3
••• the a-file, Muller 's plan is not dangerous, so the
10 �b1 ?? l"l b2+) 8 �f3 9 �et =
•.. position remains drawn.
The careless 9 �g1 ?? even loses: 9 . . . l"l g2 + ! Clearly, the problems with the realization of
1 0 � h 1 ( 1 0 �f1 l"lb2) 1 0 . . . l"lb2 1 1 a 6 �g3. Wh ite's materi al advantage were obviously
In thi s example, White 's pieces were "en­ caused by the poor position of White 's rook.
gaged in a strange role reversal ." As a rule, the Tarrasch 's famous rule i s perfectly to the point
king should support his own connected passed here : Place your rook behind the passed pawn,
pawns while the rook 's mission is to hinder the whether it 's yours or your enemy 's. Thereby the
hostile pawn. rook can retain utmost activity.
The rook's placement is extremely impor­ Tarrasch's rule is valid for the overwhelm­
tant. If the rook of the stronger side is placed ing maj ority of rook-and-pawn endings but, as it
passively (in front ofthe enemy 5 pawn) a draw goes without saying, not for absolutely all of
can be achieved simply by placing the king in them. Generally speaking, there is no rule in chess
front of the connected pair ofpawns. that has no exception.
Some time ago I was mighty impressed by a An amateur followed Tarrasch's rule in a
discovery that the ex-champion of the world pro­ correspondence game and had to resign immedi­
duced during our j oint analytical work. ately after receiving his opponent's reply. He

1 75
 Dvoretsky 's Endgame Manual

wrote an irritated letter to the grandmaster: "I Tarrasch - C higorin

relied upon your authority but lost because of St. Petersburg m (9) 1 893
you, with your stupid rule . . . "
Tarrasch published it in his chess column
with the final position of that game, adding the
following annotation: "Especially for this reader 9-92
and a few similar to him (the majority, as I am $
sure, do not need it), I supplement my rule. You
should always place your rook behind a pawn.
Except for the cases when this is unfavorable ! " B?

9-91 The game continued l . . .§ a2? 2 'it'g4 § a 1 3

§ a6+ 'it'f7 4 'it'g5 a2 5 g4! (of course, not 5 h5??
§g1 =), and Black soon resigned.
We would like to mention another, much
more complicated winning method: 2 g4 § a 1 3
.§ a6+ 'it'g7 4 h5 a2 5 'it'h2 ! ! . Just so, in order to
have, after 5 . . . 'it'h7 6 g5 §b1 7 §a7+ 'it'g8 8 .§ xa2
.§b5, the protective move 9 §g2! +- .
As Maizelis proved, the diagrammed posi­
This is a very important type of position. tion is drawn (however against a luckier setup of
The black king stands in front of connected White 's pawns, at h3 and g4, B lack has no
passed pawns but the white rook i s placed be­ chances) .
hind B lack 's passed pawn . In such cases, the 1 a21 2 h5+
•••

chances of the defender are minimal . However, If the pawn stepped ahead without giving a
th is position can be saved if Black is on move. check (e.g. with the black king on f6), the move
l. .. a 2 2 h5 �h7 (but by n o m e a n s 2 . . . § c5! would have led to an immediate draw.
2 . . . 'it'h6?? 3 §a7! 0 +- ) 3 �g5 E!g1! 4 §a7+ �g8 Well, let us make use of this idea later, when the
5 § xa2 § xg3+ white pawns reach a higher rank.
The happy end resulted from the fact that 2 �f6 3 �h4! (3 g4 § c5! 4 § xa2 'it'g5=)
•••

one of the pawns had been standing on the 3 rd 3 §h2+ 4 �g4 E!b2 5 E!a6+ �g7 6 �g5
•••

rank. l f White is on move he succeeds in advanc­ E!b5+ 7 �h4 E!b2 8 g4 (9 h6+ is threatened)
ing the pawn and wins without difficulty: 1 'it'g 5 8 �f71 9 E!a4
•••

a2 2 g4 'it'f7 3 h5 'it' g7 4 § a7+ 'it'f8 5 h6 'it'g 8 6 9 h6 .§ b6! 10 § a7+ 'it'g6= is nonsensical.

'it' g 6 - Black has no time for 6 . . . § g 1 on account 9 �g7!
•••

of the threatened checkmate. Rather than 9 . . . l"l c2? 1 0 h6 'it'g6 1 1 §a6+

If the pawn is on g2, White wins no matter 'it'h7 12 'it'h5 +- .
who is on move. He simply advances his king 10 E!a7+ ( 1 0 .§ a6 'it'f7!) 10 �f6! 1 1 g5+
..•

and the h-pawn . The riposte . . . § g 1 is useless �f5 12 h6 E!h2+ 13 �g3 E!h1 14 E! x a2
because the white rook, capturing the a2-pawn , �xg5 = .
will protect the g2-pawn .
Finally, Black has no draw agai nst the fol­ Sometimes the weaker side employs another,
lowing White setup : pawns on h3 -g4 and king more active defensive method: the king is ad­
on h4 . After l . . . a2 2 § a6 'it'h7 3 g 5 'it'g 7 3 'it'h5 vanced to support the passed pawn . As a conse­
§h1 4 § a7+ 'it'f8( g8) 5 § xa2 § xh3+ 6 'it' g 6 a quence, thi s pawn will cost a whole rook for the
winning endgame with a g-pawn ari ses . stronger side, but in the meantime his own pawns,
together with the king, will be advanced too far,
In some cases, the weaker side holds when and the endgame "two connected passed pawn
his rook protects his pawn from the side. against a rook" turns out to be winning.

1 76
 Rook Endgames

Therefore this tactic has practical chances after that move : 8 h4 a1� 9 l"\ xa 1 ®xa1 10 h5
only against less advanced pawns and misplaced l"\h2 1 1 �g5 ®b2 1 2 �h6 �c3 13 g5 �d4 1 4 g6
pieces of the stronger side. As, for example, in �e5 15 g7 l"\g2 16 �h7 �f6=.
the following case:
Tr-auic::()medies
Reshevsky - Alekhine
AVRO 1 938 Dreev - Ehlvest
USSR chsf, Tallinn 1 986

9-93
9-94

B?
B?

lt would have been an easy draw for B lack

with the white rook on a l : 1. . . �e6 2 �g3 �f6. In As we know, when the white rook is pas­
our case, however, a passive defense is hope­ sive the black king should be p l aced on the
less: l . . .�e6? 2 �g3 �f6 3 h3 �g6 (or 3 . . . l"\c2 4 queenside. However after 1 . �d5?! 2 a5 .6. �c3-
. .

�h4 l"\h2 5 l"\ a6+ �e5 6 �g5! l"\ xh3 7 l"\ xa2 l"\ h8 b3-a4 Black is very probably lost. At the proper
8 l"\e2+ �d6 9 �f6 +- ) 4 �h4 l"\ h2 5 l"' a6+ �g7 moment, the rook abandons the blockade square
6 g5 �h7 7 �g4 followed by h4, �h5 etc . g4 in order to create threats to the king. Vulfson
l �c6! 2 �g3
..• analyzed a similar endgame in detail in the book
If 2 g5 then 2 . . . l"\b5! 3 l"\ a6+ �b7 4 l"\ xa2 by Dvoretsky and Yusupov, Technique for the
l"\ xg5 5 l"\c2 l"\ gS=. The evaluation of the final Tournament Player.
position of this line is not quite obvious because It is important to push the g-pawn at least a
we have not studied defense by frontal attack single step forward in order to reduce the active
against a rook pawn . I think it is pertinent to say possibilities of the white rook.
here that, with an h2-pawn, White has winning l �f5 2 §.gl g4 3 �c4 g3?
...

chances only when the black king is cut off on But now the king fails to come back to the
the a-file. queenside in time. Black had to play 3 . . . �e6! 4
2 ... �b6 3 §.a8 �b5 4 h3 l"'g3 (in case of 4 �c5 he could resort to frontal
In case of 4 g5 both 4 . . . �b4 ( .6. l"' b3+) and checks: 4 . . . l"\c8+!?) 4 . . . �d6 5 b5 �c7 6 a5 �b7=.
the immediate 4 . . . l"\b3+ are good. 3 �d5! +- (shouldering! ) 3 ... §.d8+ 4 �c6
4 �b4 5 �f4
••. §.cS+ 5 �b7 §.gS 6 aS §.g7+ 7 �b6 �e5 8 a6
The consequences of 5 �h4 are harder to §.g6+ 9 �c5 Black resigned.
calculate, but its result is still a draw: 5 . . . �b3 6
g5 (6 �g5 l"\b1 7 h4 a1 � 8 l"' xa1 l"\ xa1 9 h5 �c4 Ostermeyer - Due ball
1 0 h6 �d5 1 1 �g6 �e5=) 6 . . . l"\b1 7 �h5 a1� 8 BRD ch, Mannheim 1 975
§ xa1 l"\ xa1 9 g6 �c4 1 0 g7 l"\g1 1 1 �h6 �d5 1 2
�h7 �e6 1 3 gS� l"\ xgS 1 4 �xgS �f5=.
s ... §.c2!
As is presumed in endgames with a far-ad­ 9-95

vanced passed pawn, Black speculates on the

threat of i n t e r fere n c e , n a m e l y 6 . . . l"\ c 4 + ,
7 . . . l"\c5(c3)+ and 8 . . . l"\ a5(a3).
6 §.bS+ �c3 7 §.aS �b4! Draw. w
Of course, there is no reason for B lack to
play 7 . . . �b2, but he seems not to be losing even

1 77
 Dvoretsky 's Endgame Manual

l �g2? Disconnected Pawns,

An odd move: in endgames, the king should One of them is Passed
go forward, not backward. 1 f3 suggested itself,
for example: l . . .b3 ( l . . . l"l c4 2 f4 or 2 l"l b6+ b. 3 If one or two files separate pawns of the
f4) 2 �f4 ( b. l"l b6+) 2 . . . l"l c4+ (side checks are stronger side, the position is most often a draw.
not efficient because the rook and the f3 -pawn We shall analyze cases of more interest and prac­
are only separated by two files) 3 �e3 l"l c3 + 4 tical value here : when the distance between
�e4 �f6 (4 . . . l"l c4+ 5 �d3 l"l f4 6 �e3) 5 f4 l"l c4+ pawns is great enough.
6 �e3 l"l c3+ 7 �d4 l"lg3 8 l"l b6+ and 9 g5 +- . The d e fe n d e r must a s p i r e fo r active
White is playing in accordance with a principle counterplay. If his rook must merely defend his
that, by Nimzovitch 's opinion, is a cornerstone own pawn or protect the king from checks, his
of a correct endgame strategy: the collective ad­ salvation is very problematic.
vance!
l ... b3 2 f4?? Miles - Webb
A severe positional error: the king will be Birmingham 1 975
cut off from the pawns forever. It was still not too
late for 2 f3! b. 3 �g3 +- .
2 �f6 3 .§b6+ �f7 4 g5 �g7 5 f5
••• =

.§c5! 6 .§b7+ �g8 7 .§b8+ 9-98

Draw. After 7 . . . �g7 8 f6+ Black can play
either 8 . . . �f7 or 8 . . . �g6 9 l"l g8+ �f7 1 0 l"l g7+
�f8.
w
Exercises

l .§a6 .§c7 2 �g5 �g7 3 f5 .§d7 4 a5

.§c7 5 .§d6!
9-96
White has improved his position to the maxi­
mum degree. Now he has in mind a typical plan
for this sort of position, a usurpation of the 7th
rank (a5-a6 and l"l d6-d8-b8-b7).
9/1 5
5 �f8 6 .§dB+ �e7 7 .§h8 �d6 8 �g6
B?
•••

.§c1 9 .§aS
A wise technique : White combines the
threat of advancing the f-pawn with an attack
against the a-pawn.
9 �e5 10 .§e8+ �f4 (10 . . . �d6 1 1 l"le6+
•..

�d7 1 2 l"l a6 +- ) 11 f6 .§gl + 12 �f7 .§a1 13

9-9 7 �g7 �f5 14 f7 .§gl+ 15 �f8 �g6 16 .§e6+
Black resigned.
I would like to draw your attention to the
fact that if the queenside pawns were placed not
9/1 6 on the same file, but on adjacent files (for ex­
W? ample, the white pawn on the b-file), the black
rook would have been less passive. It could then
combine its defensive mission with a counterat­
tacking one, and the drawing chances would
have been considerably greater.

1 78
 Rook Endgames

A typical method of bringing home a Rigan - Yandemirov

material advantage is the protection of all of Budapest 1 993
one 's pawns by the rook from the side.

Tsouros - Minev
Greece - Bulgaria m tt 1 973 9-1 00

9-99 B?

B 3 �xb5 fi xg4 4 'it'c5 .§h4! (the only method of

crossing the 6th rank with the king) 5 'it'd5 .§h6 6
.§b1 'it'g6 7 'it'e4 fih3!, the king is cut off along
the rank, and this fact is decisive.
l §dS! -+
... 2 �b2 .§c4 3 .§x b5 �f6!
Black wants to play . b6-b5; thereafter the
. . This is why the white king should have been
king, being released from its troubles with the thrown back! The rook is not hanging now, and
g5 -pawn, will set off for the queenside. White is Black manages to improve his king's position
helpless against this simple plan. Other setups without letting White do the same. If 4 .§fS+ 'it'g6
are much less efficient. 5 .§ fl , then s . . . fi xg4 6 'it'c3 fi g2! 7 'it'd3 �h5 8
2 .§f7+ (2 fi e7 b5 3 fi e8 Ei e5) 2 ... �e4 3 'it'e3 'it'g4 -+ .
.§b7 b5 4 �g4 �d4 5 �f3 4 �b3 .§xg4 5 �c3 .§e4 6 �d3 .§e8
5 �h5 �c4 6 g4 b4 7 Ei c7+ .§ cS is equally In a very similar position from the game Tal­
hopeless. l. Zaitsev (diagram 9-3 1 ), 6 . . . fi e 1 !? 7 �d2 fi e8
5 ... �c4 6 �e4 §c5 7 .§d7 b4 8 .§dl b3 was played, but in our current case Black can
9 .§bl �c3 10 .§cl + �b4 11 .§bl .§c4+ 1 2 even do without it.
�f5 g4 13 �g5 �c3 White resigned. 7 �d2
7 .§b1 is met by 7 . . . g4! 8 fibS (8 �d2 'it'g5!
If the rook protects pawns from the side 9 fi e 1 fi xe 1 1 0 'it'xe1 'it'h4) 8 ... g3 9 'it'd2 fi e4!
and the enemy king blocks the passed pawn, then 1 0 fi b3 fi g4 11 .§b1 g2 1 2 fig1 'it'g5 1 3 'it'e2
the p i e c e s of the stro n g e r s i d e attack the �h4 1 4 'it'f2 �h3 -+ 0

opponent's pawn on the other wing, while the 7 ... �g6 (7 . . . g4 is also good) 8 .§bl .§e5!
passed pawn, if necessary, can be sacrificed. (8 . . . g4?? 9 Ei e 1 =) 9 .§gl �h5 (9 . . . 'it'f5 !?) White
An interesting example of this strategy fol­ resigned.
lows in the next diagram. Studying it, we should
refresh our memories about the theory of rook As was said earl ier, only an active defense
and pawn versus rook endgames, particularly the gives the weaker side chances of salvation. We
case of frontal attack. would like to emphasize two of the most impor­
According to the above-mentioned rule, tant defensive methods:
Black must attack the g4-pawn. But how is he to 1) King 's attack against a pawn. Sometimes
do this? His king is cut off along the 6th rank one succeeds in giving the rook up for a pawn,
while l . . .fi c4? will be met by 2 .§ xb5 fi xg4 3 �c3 eating another pawn with the king and saving
'it'g6 4 �d3 .§ f4 5 .§b1 (5 �e3=) s . . . g4 (5 . . . 'it'h5 the game with a pawn against a rook.
6 'it'e3=) 6 �e3! (rather than 6 'it'e2? 'it'g5 7 .§ fl 2) Exchange ofrooks. If the eventual pawn
g3!) 6 . . . fi f5 7 'it'e2! 6 8 .§fl =. endgame is drawn, the weaker side drives away
l ... �g7!! the hosti le rook, from the rank where it is protect­
A superb waiting move that puts White in ing both pawns, by means of the exchange threat.
zugzwang. His rook is placed optimally and can­ These methods are often combined.
not abandon its place. In case of 2 'it'b4 .§ c4 +

1 79
 Dvoretsky 's Endgame Manual

Marsh all - Capablanca Tarrnanov -Averbakh

New York m (9) 1 909 Leningrad 1 94 7

9- 1 0 1 9-102

B? B

indirectly and, at the same time, protects the king

l . .§c7+ 2 \flg6 .§b7 3 h4 .§b4! 4 \flg5
from checks along files and ranks) 10 §gS a2 1 1
. .

After 4 § xb4+ ab! S aS 'it'c4 6 a6 b3 7 a7 b2

§bS+ 'it'c1 1 2 §aS §g2+ 1 3 'it'h7 'it'b1 1 4 §bS+
8 a8� b1 �+ a drawn queen-and-pawn endgame
§b2 1 S §aS §b7 -+ .
ari ses.
The move 4 §g2 is not forced but 4 §g1 a2
4 \fl x a4 5 h5 \fla3!
S § a 1 'itlb3 6 §g1 is even worse. Curiously
.•.

But, of course, not S . . 'it'bS(b3)?? 6 § xb4+

enough, Minev in the Encyclopaedia of Chess
.

ab 7 h6 +- .
Endings, annotating a similar endgame from
6 h6 .§b8 7 h7 a4 8 .§h4 .§h8 9 \flg6
Marshall - Duras, San Sebastian 1 9 1 2, evaluated
\flb3 10 \flg7 .§ xh7+ 11 \fl xh7 a3 Draw.
this position as drawn, although 6 . . . § c7! is quite
H ad Frank Marshal! been able to divine
a simple win.
his opponent's intent, he could have neutralized
The rook is a long-range piece that is ca­
it by playing 2 'it'f6(e6)! § b7 3 'it'eS ! . Here B lack
pable ofdriving the enemy king with checksfar
is in a bad way as the exchange of rooks does not
away from the decisive area. Therefore let us
work: 3 . . . § b4 4 § xb4+ ab S aS 'it'c3 6 a6 b3 7 a7
consider 4 § g4+!?.
b2 8 a8� b1 � 9 �f3+ 'it'd2 (9 . . 'it'c4 10 �c6+)
The line 4 ... '1t'c3 S §g3+ 'itld4? 6 § xa3! § xa3
.

10 �f2+ 'it'c3 1 1 �d4+ and the queens come off.

7 'it'xg7 'it'eS (7 . . . § g3+ 8 'it'f7 §h3 9 'it'g6 'it'eS
The plan of transferring the rook to b4
10 h6 �e6 1 1 'it'g7!=) 8 h6 'it'fS (8 . . . §a7+ 9 �g6!)
is nevertheless correct, however, as Igor Zaitsev
9 h7 §a7+ 10 �h6!= leads to an immediate draw.
has noted, it has to be implemented somewhat
If 4 . . . 'itlbS , then S § g S + ! CS § g2 § c7!)
differently. l . . . § c6!! l eads to the draw (prevent­
s . . .�c6 6 §g1 a2 7 § a 1
ing the white king from advancing to the center)
2 h4 § b6 (threatening 3 . . . § b4=) 3 hS § h6! 4
§ h4 'it'a3! S 'it'g7 § xhS (the rook sacrifice did not
work a move earl ier; B l ack had to wait until the
king was on g7) 6 § xhS 'it' : a4 7 'it'f6 'it'b4=. 9-1 03
The following example, as well as the exer­
cises in this section, show how difficult precise
calculation can be in this sort of position. (See
diagram, top of next column) B?
Deliberating over the natural continuation
l . . .'it'c4! over the board, Black decided that he
could get no more than a draw on account of 2
<itlg6 'it'b4 3 § g3 a3 4 § g2 a2 S § xa2 § xa2 6 What can B l ack undertake? In case of
'itlxg7=. 7 ... �cS both 8 §cl+ �b4 9 §g1 § c7! and 8 § xa2
Later on, Averbakh found an improvement: § x a 2 9 � x g 7 § g 2 + ! ( a fam i l iar too l :
4 . . . § c7! (instead of 4 . . . a2?) S 'itlh7 CS §g4+ 'itlb3 zwischenschach for gaining a tempo) 1 0 �f6 (af­
6 §g3+ 'it'b2) S . . . §cS! 6 <itlg6 (6 §h2 §gS 7 §h4+ ter 1 0 'itlh7 White will also be too late) 1 0 . . . §h2!
'it'bS) 6 . . .'1t'b3 7 § g3+ 'it'b2 8 §g2+ §c2 (this is 11 �g6 �d6 12 h6 �e7 13 h7 'itlf8 are bad.
why the zwischenzug S . . . §cS! was necessary - Summing up: a rook sacrifice for the a-pawn holds
the king prevents a rook capture on g7) 9 §g1 when the black king is on d4, c4, or b5, but not on
§h2! (from here, the rook defends the g7-pawn c5 or c6.

1 80
 Rook Endgames

Let us try the waiting move 8 �h7 ! . Now

8 . . . �c4 9 l"' xa2!= is useless; 8 . . . g5+ 9 �g6 g4 1 0
h 6 g 3 1 1 h7 l"\ xh7 1 2 �xh7 g 2 is not dangerous
for White either, because the black pawns are
too far away from each other (look at diagram 8-
33 again).
8 . . . �d5 is the strongest. After 9 �g6, Black
does not play 9 . . . �e5 10 l"\ e 1 +! ( 1 0 �h7? �f5)
10 . . . �d4 1 1 l"\ a 1 ! /::;. 1 2 l"' xa2 (the king is too late
approaching the pawn). Instead, he has 9 . . . �d6! .
This position could have been reached 2 moves 5 §g4+ �b3 6 §g3+ �c2 7 §g2+ �d3 8
earlier, if Black played 7 . . . �d6! (instead of §g3+ �c4 9 §g4+ �b5 10 §xg7??
7 . . . �c5). The elementary 10 �d4! led to an immedi­
At first I did not see any danger for White ate draw. The capture is much weaker because
here as well: 10 �h7! ( 1 0 l"\ d 1 +? is bad because Black maintains the possibility of interference
of 10 . . . �e7! 10 l"\ a 1 �f8 followed by . . . l"\ a6+) with his rook along the 6th rank.
1 0 . . . �e5 1 1 �g6. However, grandmaster Muller 10 ... a3! 11 §a7 (1 1 l"\ g 1 l"\ xh5+ 1 2 �d4
finally discovered a winning continuation. Black �b4 1 3 l"' b 1 + �a4 14 �c3 a2 1 5 l"\ g 1 �a3 -+ )
suddenly sacrifices his g-pawn : 1 1 . . . l"\ a6+ ! ! 1 2 ll . §a6 12 §b7+
..

�xg7 �f5 1 3 �f7 ( 1 3 h 6 l"\ a7+) 1 3 . . . �g5 1 4 1 2 l"\ g7 a2 1 3 l"\ g 1 did not help : 1 3 . . . l"\ h6!
�e7 �xh5 and, as can easily be seen, his king 1 4 l"\ a 1 (14 �d4 �b4) 1 4 . . . l"\ xh5+ 15 �d4 l"\ h2
comes to the queenside in time. 1 6 �c3 �a4 -+ .
1 §a6?! (An attempt to cut the king off
.•.
12 ... �a4 13 §g7 §a5+ 14 �f6 a2 15
from the g7-pawn does not work, although it does §g4+ �b3 (15 ... �b5?? 1 6 l"\ g 1 =) 16 §g3+
not spoil anything as well). �c4! 17 §g4+ �d3 18 §g3+ �e4 19 §g4+
2 �f5 �c4 3 §g3! �e3 Cl 9 . . . �f3) 20 §g1 § xh5 21 §g3+ �d4
This is the point! The line 3 . . . a3 4 l"\ xg7 a2 22 §a3 §h2 23 �f5 §f2+ 24 �g4 �c4, and
5 l"\g1 leads only to a draw.
White resigned soon.
3 ... §f6+
Black should have played 3 . . . l"\ a7! 4 �g6 Exercises
�b4 (rather than 4 . . . a3? 5 l"\ xa3) 5 l"\ g4+ �c5 ! In both cases, your task is to find whether
(5 . . . �b5 6 l"' g3!) 6 l"\g5+ (6 l''lx a 4 l"\ xa4 7 �xg7 Black can achieve a draw.
l"\ g4+!) 6 . . . �c6(d6), transposing into situations
that are already familiar to us. For example: 7 l"\g2
(7 l"\g1 a3) 7 . . . a3 8 l"\a2 (8 l"\ c2+ �b5 9 l"\g2
l"\ c7!) 8 . . . �c5 ! 9 l"\ a 1 a2 1 0 �h7 �d5 1 1 �g6 9-1 05
�d6! etc. "a la Muller."
4 �e5?!
In spite ofAverbakh's opinion, 4 �g5 gives
no draw. Black should simply return with his rook
to a6 (see the previous annotation). 911 7
Averbakh 's line 4 . . . �b4 5 l"\ g4+ �b3?! B?
(5 . . . �b5 ! 6 l"\ g3 l"\ a6 7 �f5 l"\ a7 8 �g6 �b4 9
l"\ g4+ �c5! -+ ) 6 l"\g3+ �c2 7. l"\ g2+ �cl 8 l"\g3
l"\ a6? (8 . . . �b2! 9 l"\g2+ �b3 1 0 l"\ g3+ �b4 1 1
l"\ g4+ �b5 -+ ) 9 �f5 l"\ a7 actually leads to a
draw: 1 0 �g6 �b2 , and now 1 1 �h7! a3 1 2 9-106
l"\ xg7= rather than 1 1 l"\g2+? �b3 etc. (See next
diagram)
4... §h6??
As is known, the one who wins errs next to
9/1 8
last (White's decisive error is still to come). Black
should have played 4 . . . l"\f7! 5 l"\ g4+ �b5 6 l"\g3 B?
l"\ a7 7 �f5 a3 8 �g6 a2 9 l"\ g 1 �c4 etc.

181
 Dvoretsky 's Endgame Manual

Four Pawns vs. Three on the Same Wing

I f all pawns are on the same wing, bringing ct;f6 20 §a2 ctig7 21 §e2 ct;f6 22 §e3 ct;g7
the advantage home is frequently impossible (it 23 e5 ct;f8 24 g4
is more precise to say, it should not be possible lf 24 �g5 , the most simple is 24 . . . �g7, al­
against correct defense ) . The fewer pawns, the though 24 . . . l"\ xg3+ 25 �f6 �g8 26 l"i d3 l"i h3 27
easier the defense is. e6 fe 28 �xg6 l"i g3+ 29 �xh5 �g7= or 27 l"i d8+
Say, with 3 pawns against 2 or even with 4 �h7 28 �xf7 l"\ xf3+ 29 �e7 g5 30 hg h4 31 e6
against 3 , in case of standard pawn structures, h3 32 l"\ d2 �g6= is also playable.
the task of the defender is not too difficult (once 24 ...hg 25 fg ct;g7 26 ct;g5 §fl 27 §e4
in a lightning tournament I managed to hold two §f3 28 h5 (28 e6?? f6 # ) 28 ... gh 29 gh f6+ 30
such endings : against Tal and Vasiukov) . As for ct;g4
the case of five pawns against four, the probabil­ Or 30 ef+ l"\ xf6 3 1 l"ie7+ l"if7 32 h6+ �g8=.
ity of losing is rather great. 30... §f1 31 h6+
A l ittl e trap before the curtain fal l s .
Petrosian - Keres 3 l . . .�xh6? loses to 32 e6 f5+ (if32 . . . l"\ g l + , then
USSR eh, Moscow 1 95 1 either 33 �f4 l"\ g8 34 �f5 or 33 �f5 l"\fl + 34
l"\ f4 l'=! xf4+ 35 �xf4 �g6 36 �e4) 3 3 �h3! fe
34 e7.
31 ... ct;g6! Draw.
9-1 0 7 As can be seen, Black did not have serious
troubles.
It should be mentioned that, when the white
pawns had been set into motion, Keres used a
B typical strategic policy for this sort of position :
attacking the pawns from the rear.

What if B lack could not play . . h7-h5 in .

1 . . . h5! time? We shal l analyze two important endings

In this way Black makes his task of reaching that may serve as landmarks for both sides: the
a draw considerably easier. The defender should stronger side may pursue them while the weaker
advance his h-pawn. The stronger side, when­ side should avoid these situations.
ever possible should prevent this by means of These endings are thoroughly analyzed in
g3 -g4L endgame handbooks. We skip some less impor­
The explanation consists in the fact that tant lines but bring respective conclusions.
White 's most logical plan is an advance of his e­
and f-pawns in order to create a passed pawn . To Botvinnik - N aj dorf

accomplish this plan, he must sooner or later play Moscow ol 1 95 6

g3-g4, allowing a pawn exchange on g4. But, as
we know, pawn exchanges are usually favor­
ablefor the weaker side, and improve the draw­
ing chances. Without . . h7-h 5 , the h-pawns
.
9-1 08
would have stayed on the board .
In this game, Petrosian gradually carried out
another plan : h2-h4 followed with f2-f3 and g3-
g4, but also could not obtain victory. w
2 §c2 ct;g7 3 ct;g2 §b5 4 ctif3 ct;f6 5 h4
§f5+ 6 ct;g2 §a5 7 ctih3 §a4 8 §d2 ctie5 9
§b2 ct;f6 10 §b5 §a2 11 ct;g2 §a4 12 ctif3
§a3 13 ct;f4 §a2 14 f3 §e2 15 e4 §e1 16 1 § a 5 §c7 2 §d5 §a7 3 e 5 fe 4 fe ( 5
§b6+ ct;g7 17 §a6 §b1 18 §c6 §g1 19 §c2 l'=l d7+! is threatened) 4 . . . ct;e7 5 e 6 §a4!

1 82
 Rook Endgames

5 . . . B a6 6 B d7+ \t'f8 7 \t'g6! B xe6+ 8 \t'h7 is Capablanca - Yates

quite bad for B lack. Hastings 1 930/3 1
6 g5!
6 B d7+ \t'f8 7 Bf7+ \t'g8 8 g5 fails in view
of 8 . . . B a 5 + ! (8 . . . hg? 9 \t'g6) 9 \t'e4 B a6=.
6 ... §a7! 9-11 0
The best defense as suggested by Aronin.
The rook may return because there is no danger
of trading the rooks anymore : 7 B d7+? B xd7 8
ed 'it'xd7 9 \t'g6 hg 1 0 'it'xg7 g4= . w
The actual continuation was 6 . . . hg?! 7 B d7+
\t'f8 8 Bf7+ \t'g8 9 \t'g6 g4 10 h6! (the shortest
way to a win) lO . . . gh 1 1 e7 B a8 1 2 B f6 ( D. B d6-
d8) B lack resigned. instructive errors on the way to the final out­
7 §e5! come.
A key move ! White protects the pawn and l §b6?
prepares a king invasion. White should have played 1 B d6! in order
An anticipatory pawn exchange is errone­ to use the rook to protect against checks from
ous : 7 gh? gh 8 Bb5 ( D. B b6) 8 . . . B c7! 9 Bb6 the side. The correct reply to the move actually
Bc5+! 1 0 �g6 B e5! 1 1 �xh6 \t'f6! 1 2 B a6! Bf5 ! . played in the game was 1 . . . B a4! 2 \t'f3 (2 'it'g3
B a3+ 3 �h4 B a4 4 f5 Ba5 5 e6 fe 6 fe \t'f6=)
2 . . . Ba3+ 3 \t'e4 Ba4+ 4 \t'f5 B c4 5 Bb7 ( D. 6 e6)
5 . . . \t'f8.

9-111

w
I n this position, Black must play very pre­
cisely in order to achieve a draw, but theory says
that this goal is within his reach.
7 ... hg White m i s sed the correct way : he has
7 . . . \t'd6 8 gh gh 9 \t'f6; 7 . . . B a6 8 \t'g6 'it'f8 brought his king, not his pawn, to f5 , so he can­
9 \t'h7 hg 1 0 e7+ \t'e8 1 1 \t'xg7 g4 1 2 h6 +- . not win anymore.
8 <if} xg5 (8 \t'g6 \t'd6 9 B e l g4 1 0 h6! gh l §e3? 2 §b4
••.

1 1 \t'f6 is also strong) s . . §al 9 <if}g6 §fl

. 2 B b8 suggested itself, however after
(9 . . B g l + 1 0 B g5) 10 <if} xg7 §gl + 11 <if}h6!
. 2 . . . B e4 3 �f3 B e l a straightforward 4 B e8?
§g2 12 §g5 +- . enables the salvation through 4 . . . h5! 5 g5 (5
In the next diagram, White 's position is win­ gh B fl +! 6 \t'e4 B e l + 7 \t'f5 B h l ) 5 . . . B fl + 6
ning (the same evaluation is valid with the black 'it'e3 h4.
pawn on h7 and the white pawn on g 5 ) . The The most precise is 2 Bbl ! (temporarily de­
winning plan i s a rook trans fer to the 8th rank nying the black rook the I st rank). Black is in
followed by f4-f5 -f6+. If the black rook aims at zugzwang. He must either worsen his king's po­
the e5 -pawn, Wh ite defends it with the rook sition or move his rook off the e-file where it is
from e8. best placed. In both cases, the invasion of the
Capablanca carried this plan through; how­ white rook gains in effectiveness. For example,
ever, as renowned rook endgame expert Kopaev 2 . . . B e4 (2 . . . B e2+ 3 \t'f3 Bh2 4 f5 h5 5 Bb7 hg+
demonstrated, the opponents made a number of 6 'it'g3 Bh5 7 \t'xg4 Bhl 8 e6 +- ) 3 'it'f3 B a4,

1 83
 Dvoretsky 's Endgame Manual

and now time has come for the main plan : 4 .§. b8! before the position of the diagram 9- 1 1 0 arose,
.§.a3+ 5 �g2 .§. e3 6 .§. e8! .§.e2+ 7 �f3 .§.e1 8 f5 and Yates could have had an easy draw by means
.§. fl + 9 �e2 .§. f4 1 0 �e3 .§. xg4 1 1 f6+ �h7 1 2 of . . . f7-f6 .
e6 +- . 2 ... .§.a7?
2 E!c3 3 ct;f2? (3 .§. b8) 3 ... .§.a3?
.•• Antoshin had to keep in mind the danger of
Both adversaries missed the fact that after a check along the 8th rank: his king, when stand­
3 . . . h5! Black either trades a pair of pawns (4 gh ing on h7, is too far removed from a passed e­
.§.h3) or (in case of 4 g5 h4) obtains enough pawn if White manages to create it. Therefore
counterplay to save the game. here, and later on too, he should have played f7-
4 .§.b7?1 (4 .§. b8!) 4 . ct;g8?1 (4 . . . .§. a2+!?) 5
. . f6 ! . Black could then parry the threat oftranspo­
.§.b8+! ct;g7 6 f5 ( 1:::, 7 f6+ ) , and White won. s i t i o n i nto the w i n n i n g p o s i t i o n fro m the
Botvinnik-Naj dorf game by means of catching
Korchnoi - Antoshin the white pawns from behind. A characteristic
USSR eh, Erevan 1 954 variation was demonstrated by Korchnoi : 2 . . .f6!
3 .§. c8+ �f7 4 .§. c7+ �g8! 5 �f3 (5 f4 .§. a3 -

White 's king is cut off from his pawns) 5 . . . .§.a3+

6 �f4 .§.a2 7 �f5 �h7! (this is why the black
9-112 king drew back to g8) 8 f4 .§. f2 ! = as White fails to
create the passed e-pawn.
It is time to explain why the move l . . . .§. a5
was given an exclamation mark. With a pawn on
W? g3, this defensive plan does not work : the f4-
pawn is protected, so White can play �e6, while
Black can hardly prevent the penetration of the
l h51 white king to f5 (via g4) at an earlier stage.
B lack, if he was on move, could have con­ 3 .§.c6?
siderably simplified his task by placing his own After 3 .§. c8 + ! �h7 4 e 5 ! ( 1:::, �g3 , f4 ,
pawn to h5. If 1 g4?!, then all the same 1 . . . h5 ! . .§.e8 +- ) Black would have been faced with prob­
l.. . .§.a51 lems one can hardly tackle over the board.
As Korchnoi noted in his exceptionally deep
and far-reaching comments to thi s endgame, it is
useful for Black to force the advance g3-g4.
2 g4 9-114
2 .§. c8+ �h7 3 g4 is not dangerous yet be­
cause of 3 . . . g5! 4 hg+ �xg6 ( 1:::, 5 . . . h5) 5 f4 f6! .

9-113

As Korchnoi showed, almost all defensive

methods are doomed to lose : White either cre­
w ates a dangerous passed e-pawn or transposes
to positions from the game Capablanca-Yates.
For example, 4 . . . g5? 5 hg+ �xg6 6 f4 +- , or
4 . . . .§. a4?! 5 �g3 .§. e4 6 .§. e8 g5 (6 . . . .§. a4 7 .§. f8
The last move is worth special attention. It .§. a7 8 f4 +- ; 6 . . .f6 7 e6 g5 8 hg+ �xg6 9 f3 .§. e 1
is vitally important for Black to prevent the 1 0 �f2 .§. e 5 1 1 f4 +- ) 7 hg+ �xg6 8 f3! + - ( 8 f4?
pressing advance e4-e5 that leads to the setup is not precise, Black holds after 8 . . . h5!). Of course,
from the Capablanca-Yates game. By the way, in only basic results are shown here, as a detailed
that game the white pawn stood on e4 a few moves explanation would have been rather complicated

1 84
 Rook Endgames

and too vast. 1 2 . . . fe+ 1 3 fe El xg4+ 1 4 �d5 El g 1 1 5 e6 El d 1 +

4 . . . El e7! (the only defense) 5 f4 f6! 6 El c5 fe 1 6 �c6 El e 1 1 7 �d7 El d l + 1 8 �e8 +- is no
7 fe, and now Black must prevent the white king's better.
march to the center, that would transpose to the 13 e6 §.xf4+ 14 �d5 E!f5+ 15 �d6 §.xh5
B otvinnik-Naj dorf ending, by 7 . . . El f7 ! . Here 16 e7 E!e5 17 e8� E! xe8 18 E!xe8
Kor:chnoi gives 8 �g3 g6!= and 8 e6 El e7 9 E\ c6 The fight is almost over. When the white
g6 1 0 El d6 ( .c:,. 1 1 El d7) 1 0 . . . Ele8 1 1 �f3 gh 1 2 king comes back to his home side of the board,
gh �g7 1 3 �e4 �f6 1 4 �d5 El e7=. the rook will be stronger than 3 pawns.
The prophylactic move 8 Eld5! is more dan­ 18 �g6 19 �d5 �f5!? 20 §.e1
•••

gerous for Black. He cannot play 8 . . . g6? on ac­ 20 �d4 �f4 21 �d3 �f3 22 Elg8 g5 23
count of 9 e6 El f8 (9 . . . El e7 10 El d7) 10 El d7+ El f8 +- is also strong.
�g8 1 1 e7 El e8 1 2 El d8 �f7 1 3 El xe8 �xe8 1 4 20 h5 21 E!fl+ �g4 22 �e4 g5 23 E!xf6
.••

g 5 ! +- , and 8 . . . g5? 9 e 6 E\ f8 1 0 e 7 El e8 1 1 El e 5 h4 24 �e3 �g3 25 �e2 g4 26 �fl �h2 27

�g7 1 2 El e6 +- i s also bad. Therefore h e must §.f4 h3 28 §. x g4 �h1 29 �f2 h2 30 �g3
wait: 8 . . . Elf8 9 �g3 Elfl , and if 10 El d3, then �g1 31 �h3+ Black resigned.
10 . . . g5 ! . But I doubt whether Black can hold this
endgame after 10 El d7! 0 El f8 (the same reply Tr-auic()medies
follows to 10 . . . �g8) 1 1 El d3 g5 1 2 Elf3 El e8 1 3
El f5 followed by 14 �f3 . The two last endings fully fit this category,
3 §.a3? (3 . . .f6!=) 4 f3?
••.
but I would like to add some new examples, the
4 El c8+! �h7 5 e5! was winning. last of which has some theoretical value.
4 §.a51? (4 . . .f6!)
•••

Now White can gradually strengthen his Bellon - Chekhov

position by means of El c8-d8, �g3-f4 or f3-f4, Barcelona 1 984
but, as his pawn cannot come to e5, the game will
be drawn if Black defends precisely.
Korchnoi decided to force the events and
was successful, but only due to a new mistake 9-116
by Black.
5 §.c8+ �h7 6 f4?1 ( ..6. 7 e5 +- ) 6 ... §.a2+
7 �f3 E!a3+ 8 �f2 §.a2+ 9 �e3 E!a3+ 10
�d4 w

9-11 5
The waiting policy (1 Ela5 or 1 Elb7) gave a
rather easy draw, but Bell6n decided to chase
after the g7 -pawn.
1 E!b8+ �e7 2 §.g8?? §.d8!
B?
White resigned. The pawn endgame is quite
hopeless for him, while after 3 El xg7 his rook is
lost: 3 . . . �f8 4 El h7 �g8 5 El xh6 �g7 6 �g5
Eld5+.
Black can hold the game rather simply :
1 0 . . . El f3! 1 1 �e5 (1 1 f5 f6=) 1 l . . .f6+ 1 2 �f5
El fl , achieving the position from the note to
Black's move 2.
1 0 §.g3?
.••

He chases after material gain but lets White

create a passed pawn that will cost him a rook.
11 E!f81 f6 12 e51 §. xg4

1 85
 Dvoretsky 's Endgame Manual

Norri - Svidler gf does not lose) 7 f5 gf 8 e6 (8 �xh5 �f8 !:>.

Erevan ol 1 996 9 . . l"\ xe5) 8 . . .f4+! 9 �xf4 �f6 1 0 .§ xt7+ �xe6=.
.

l §d3?1
.••

In many similar situations, to place the rook

behind the e-pawn makes some sense; particu­
9-11 7 larly, such a maneuver is not bad when h-pawns
are absent. But here this transfer is erroneous.
Its slightly modified version does not work, ei­
ther: l . . . l"' d4?! 2 �h3 .§ e4? (in case of 2 . . . g5?
w White does not play 3 fg �g6, he has 3 l"' c7!
instead) 3 l"' c7! (3 �h4 �h6 !:>. 4 ... g5+) 3 ... �g7
4 �h4 .§e2 5 �g5 ! , and we come to situations
that have actually occurred in the game.
Here again White did not have enough 2 �h3 §e3?
patience and made an analogous error. 2 . . . l"' d7 3 �h4 �h6= was necessary.
1 f5? ( 1 l"\a7=) l . .. gf 2 §h7 �g8 3 3 �h4?!
§xh5?? Playing 3 l"' c7! �g7 4 �h4, Piket could have
3 l"' a7 could still hold the game. chained the hostile rook to the e-file and, as we
3 §e5 4 �f3 f41 White resigned.
••.
shall see, this was a winning method.
3 �g7?
Piket - Kasparov
••.

He should have tried 3 . . . �h6! 4 l"' c7 l"' e 2 ! .

An Internet Tournament, 2000
If 5 �h3, then 5 . . . �g7 (5 . . . g 5 ! ? is also playable)
6 l"' b7 g5! 7 fg �g6. The line 5 g4 hg 6 .§ xt7
.§ xh2+ 7 �xg4 is more dangerous for Black, but
9-11 8
after 7 . . . l"'e2 he seems to be surviving.
4 �g5?
$
An erroneous order of moves, again 4 l"' c7!
l"'e2 5 �g5 is correct. Now Black could return to
B Averbakh's plan : 4 . . . l"'a3! 5 l"' c7 l"' a5=. However
Piket could hardly expect that his opponent would
suddenly change his mind and move the rook
back.
White succeeded in bringing his pawn to 4 ... §el? 5 §c7 §e2 6 §e71 §a2 7 f51 gf
8 e6 h4 9 §xf7+ �g8 10 �f6 Black resigned.
e5 (generally speaking, it would have been fa­
Let us look at 6 . . . l"'e4 (instead of6 . . . l"' a2).
vorable for Black if he prevented this by playing
. . . t7-f6 at an earlier stage). On the other hand,
. . . h6-h5 is already played, so reaching a draw
should not be a very difficult problem.
Kasparov had to decide how to behave in 9-11 9
case of the white king's march to g5 via h3 and $
h4. The simplest method was to play . . . �h6 at a
proper moment. For example l . . .�g7 2 �h3 l"' a7
3 �h4 �h6!, and one cannot see how White W?
could make any progress.
Moreover, a king invasion to g5 is not too
dangerous. Even with the white rook on the 7th
rank Black can survive. Averbakh analyzes 3 . . . l"'a6 This position occurred in the following
(instead of 3 . . . �h6) 4 l"' c7 .§ b6 5 l"' e7 .§ a6 6 games: Stean-Hartston (Great Britain eh, Brighton
�g5 (6 e6 �f6! 7 .§ xt7+ �xe6=) 6 . . . l"' a5 ! (as 1 972), Ionov-Karasev (Leningrad 1 9 83) and
Bologan says, even 6 . . . l"' b6!? 7 e6 .§b5+! or 7 f5 Matveeva-Rappoport (Baku 1 983). In all these
games, White found a forced win.

1 86
 Rook Endgames

7 e6! .§.xe6 8 .§. xe6 fe 9 h3 �f7 10 �h6 �f6 White must wait: 6 l"\ b7! 0 l"\ a4 (6 . . . l"\e2 7
ll g4 h4 (1 1 . . . hg 1 2 hg O +- ) 12 g5+ ( 1 2 <it'h7? l"l e7 ! transposes to the actual course of the
g5) 12 ... �f5 13 �g7 �xf4 14 � xg6 e5 15 game), and here Bologan has discovered a bril­
�f6! e4 16 g6 e3 17 g7 e2 18 g8� e1� 19 liant solution : a double pawn sacrifice 7 g4! !
�g4+ �e3 20 �e6+ �f2 21 �xe1 + �xe1 22 h g ( 7 . . . l"\ e4 8 g h g h 9 h 4 is hopeless) 8 f5 ! gf 9
�g5+ - e6 +- .
Finally, instead of 5 . . J'le2 Black could have A gain of another pawn is much weaker: 7
played 5 . . J'\ e4!? at once. The point is to meet 6 e6 l"\ a 5 + 8 <it'h4 �f6 9 ef <it'g7 1 0 l"\ e7 �f8! 1 1
l"l e7 with 6 . . . l"\ a4!, and 7 f5, as was played by l"l e 5 l"\ a 2 , or 1 0 h3 <it'fS 1 1 g4 hg 1 2 hg l"\ c5 1 3
Piket, is not possible anymore, while if7 e6, then f5 l"l c6! (rather than 1 3 . . . gf? 1 4 g5+-). I t looks
7 . . . l"\ a 5 + 8 <it'h4 <it'f6 9 l"l xf7+ <it'xe6=. like Black holds in both these lines.

Balance on One Wing and an Extra Pawn on Another

Situations with an extra remote passed pawn faced with more complicated problems . After a
occur now and then, therefore it is very impor­ king's march to the queenside Black removes his
tant to learn their correct evaluation and han­ rook from b6 either for protecting his own pawns
dling. The decisivefactor in this sort ofendgame or for attacking the hostile ones.
is the position of the rook of the stronger side. 2 b5 �e6 3 b6 )3c8 4 h3
In majority of cases the rook is placed best "a 4 b7? l"\ b8 is erroneous because it allows
la Tarrasch, " behind its own passedpawn; some­ Black to eliminate the b-pawn and thereafter to
times its sideways position is preferable. bring his king back to the kingside in time. For
example, 5 ®g1 ®d6 6 ®f2 ®c6 7 ®e3 l"l xb7 8
Quite often, however, we Jack free choice,
l"\ xb7 ®xb7 9 �e4 ®c6 1 0 ®e5 ®d7=.
so the rook mostly stands in front of the pawn in
4 ... )3b8 5 �h2 �d5
practical games. Therefore we shall pay more at­
l f the black king stays with his pawns, his
tention to these cases.
adversary heads to the b-pawn. Black cannot pre­
vent this by means of the opposition because
The Rook Behind its Own Pawn White can make a waiting rook move; Black will
then be obliged to give way to the white king
Botvinnik - Boleslavsky
because his rook has no waiting moves. This
Leningrad/Moscow 1 94 1 clearly demonstrates the difference between the
rook positions.
6 �g3 �c6 7 �g4 �b7
A capture on b6 is impossible now; there­
9-120 fore Black blocks the pawn with his king, re­
leasing the rook from this duty. A standard and
often quite useful method ; but alas, it does not
bring any relief to B l ack in this particular case.
W? 8 l3e1!
Excellently played ! While the rook was pin­
ning the black rook down it was superbly placed
on b I , but now it will be more active when placed
1 )3b1! sideways. In case of8 . . . \tlxb6 9 l"lb1 + Black loses
The rook has occupied its correct position the pawn endgame.
behind the pawn. After 1 h3? l"lb2! 2 l"\ e4 Black 8 ... )3g8 9 l3e6 �a6 10 �g5 �b7 11 h4
could have achieved a draw. The rest is simple. White attacks on the
1 ... �f7? kingside, having an extra piece there.
The passed pawn should be blocked as ll �a6 12 h5 �b7 13 g4 �a6 14 �h4
••.

soon as possible. Black had to play 1 . . . l"l c6! 2 b5 �b7 15 h6 gh 16 .§. xh6 .§.g7 17 �h5
l"\ b6. I do not think this was enough for a draw ( .6. g5, l"l e6, �h6 +- ) 17 . . �a6 18 )3c6 .§.e7
.

but, anyway, his opponent would have then been 19 .§.c7 .§.e5+ 20 g5 �xb6 21 .§.xh7 �c6 22

1 87
 Dvoretsky 's Endgame Manual

�h6 �d6 23 g6 .§el 24 .§fi �e6 25 .§f2 .§al The exchange of pawns makes White's task
26 g7 §.hl + 27 �g6 §.gl + 28 �h7 §.hl + 29 easier. As was revealed in later analyses, after
�g8 �e7 30 §.e2+ �d7 31 §.e4 ("bridging") 3 . . . 'it'e6! Black would sti l l have had winning
3l ... §.h2 32 �fi Black resigned. chances .
4 fg+ �e5
Botvinnik - Borisenko In case of 4 . . . 'it'g5 !? White simply waits : 5
U S S R eh, Moscow 1 955 'it'g2 'it'h4 6 'it'g1 'it'h3 7 'it'h1 l"l. e7 (7 . . . f5 8 gf g f
9 'it'g1 f4 1 0 'it'f2=) 8 l"l.a3+ 'it'xg4 9 l"l. xa5. Kopaev,
as well as Levenfi sh and Smyslov, evaluate th is
position as drawn although after 9 . . . 'it'f3 ! this is
9-1 2 1 far from obviou s . Instead of 6 'it' g l , Botvinnik
recommended 6 h3!?; and Marin proved that it is
indeed enough for a draw : 6 . . . g5 7 'it'h2 l"l. b7 8
l"l. xa5 l"l. b2+ 9 'it'gl 'it'xh3 1 0 l"l. a6 'it'xg4 1 1 l"l. xf6
w 'it'g3 1 2 l"l. fl ! = .
5 h4 �d5 6 h5 gh 7 gh
The goal is reached; White has created a
passed pawn. Black cannot win anymore, for ex­
1 §.a4! ample 7 . . . 'it'c5 8 h6 'it'b5 9 l"l. h4 l"l. h7 10 l"l. h 5 +
Botvinnik blocks the pawn immediately. I f 'it'b4 1 1 l"l. h4+ 'it'b3 1 2 l"l. h3+ 'it' b 2 1 3 l"l. h4=.
h e al lowed . . . a5-a4 h e would have had n o chances 7 . . . �e6 8 h6 �f7 9 §.g4! �f8 10 §.f4
at a l l . A l ekhine won a s i m i lar ending from §.a6 11 §.g4 §.a7 12 §.f4 �g8 13 §.xf6 a4 14
Capablanca in the last, 34th, game of their match §.f2 �h7 15 §.a2 �xh6 16 �f2 �g5 17 �e3
for the World Championship in 1 92 7 : it can be Draw.
found in almost every book on endgames.
l .. �g5?
.
T.-aeic:;()medies
An instructive error. The king heads for the
queenside, but a safer road was via g7. Why? Dvoretsky - Kupreichik
The point is that the best chancefor a successful USSR ch( l ), Minsk 1 976
defense in this sort of position is counterplay
on the kingside: creation of a passed pawn or
weakening the opponent :� position. The posi­
9-122
tion of the king in front of the pawns contributes,
as we shall see, to the adversary 's counterplay.
After l . . .'it>g7! White is not getting on:
2 f3 �f7 3 g4 h4 -+
2 �g2 �f7 3 �f3 'it'e6 4 h4 (4 g4 h4 5 g5 B?
fg 6 'it'g4 'it'f6 7 h3 l"l. a8 -+ ) 4 .. .f5 5 'it'f4 'it'd5 6
�g5 l"l. a6 7 f3 'it'c5 8 g4 fg 9 fg hg 10 'it'xg4 'it'b5
l l l"l. a 1 a4 -+ (Levenfish, Smyslov)
The di agrammed position aro se in an
2 h4 'it'f7 3 'it'fl 'it'e6 4 'it'e2 'it'd6 5 'it'd3 (5
adjourned game a few moves after resumption of
g4 hg 6 l"l. xg4 a4 7 l"l. xg6 a3 8 l"l.g1 a2 9 l"l. a 1 'it'e5
play, so both the adversaries had reached it in
10 'it'f3 l"l. a4 -+ ) 5 .. .f5! (5 . . . 'it'c6? is erroneous in
their home analyses.
view of 6 g4 l"l. d7+ 7 'it'c3 l"l. d5 8 l"l.f4 f5 9 gh gh
I only expected a logical maneuver that
10 'it'c4=) 6 f3 'it'c5 7 g4 'it'b5 8 l"l. d4 a4 9 �c2 a3
p l a c e d the rook behind the p a s s e d paw n :
10 'it'b1 l"l. a4! l l l"l.d6 hg 1 2 l"l. xg6 gf-+ (Kopaev).
l . . .l"l.h7+! 2 'it'g3 l"l. h l . I n that case, after 3 'it'f4
2 f3! �f5
l"l.a1? 4 g5! fg+ 5 'it'xg5 l"l.a3 6 l"l.f4 Black was lost,
After 2 . . . f5!? 3 'it'f2 'it'f6 4 h4 'it'e5 5 �e3
but 3 . . . l"l. f1 ! , preparing . . . g6-g 5 + , destroyed
'it'd5 6 g4! the outcome is also unclear.
White's plan.
3 g4+! hg?
l §.b7?!
..•

1 88
 Rook Endgames

A peculiar move: my rook can occupy a po­

sition behind the pawn now, and so even in two
ways: 2 f! e3 6 3 f! a3 and 2 aS 6 3 f! a4.
The second way is apparently more attrac­ 9-124
tive : in principle, it is favorable to push the pawn
farther. So I stepped into it, failing to discover a
cleverly prepared trap. The correct continua­
tion was 2 f! e3 ! f! a7 3 f! a3 f! a 5 ! 4 �g3 and W?
Black's position is still very difficult, very prob­
ably lost.
2 a5? §b31 3 §a4 § xf3 4 a6
It seems so that the pawn can only be
stopped by means of 4 . . . g5+ 5 �h5 f! h3+ 6 �g6 6 ®f6?
f! h8 7 a7 f! a8, and this is surely hopeless for A decisive loss of a tempo ! Lasker saw the
Black. correct way but, as he explained after the game,
4 ... <itle61! he wanted instinctively to avoid a discovered
It comes to light that after 5 a7? g5+ 6 �h5 check along the 7th rank.
�f7 White will be checkmated. He should have performed the breakthrough
5 g5 fg+ 6 ® xg5 §fS Draw. that gave him a passed pawn immediately: 6 f5 !
ef (6 . . . gf 7 h5) 7 e6 fe+ 8 �xg6. After 8 . . . �b5 9
Em. Lasker - Levenfish f! a 1 f4 1 0 h5, both 1 0 . . .f3 1 1 f! f1 a4 1 2 f! xf3 a3
Moscow 1 925 1 3 f! f1 a2 1 4 f! a 1 �c4 1 5 h6 �b3 1 6 h7 f! a8 1 7
f! e 1 ! e 5 ( 1 7 . . . �b2 1 8 f! e2+) 1 8 �g7 e4 1 9 h8�
f! xh8 20 �xh8 (the black pawn cannot reach e2,
so there is no win) and 10 . . . e5 1 1 f! e 1 ! �c4
9-123 (l l . . .a4 1 2 f! xe5+ �c6 1 3 f!e4 a3 14 f! xf4 a2 1 5
f! fl =) 1 2 f! xe 5 �d3 1 3 h 6 f3 1 4 h 7 f! xh7
(14 . . . f! a8 1 5 f! xa5) 15 �xh7 f2 16 ms �e3 1 7
f! fS (or 1 7 f!e 5 + �f4 1 8 f! e8) 1 7 . . . a 4 1 8 f! e8+
w �f3 1 9 f! fS+ �g2 20 f! g8+ �h3 21 f! fS= lead to
a draw.
6 <itlb5 7 §a1 a4 8 f5 ef 9 e6 fe 10 ®xg6
.•.

f41 11 h5 f31 12 h6 ( 1 2 f!fl a3 -+ ) 12 e5! 13

1 g4
•..

§e1
L asker aspires for counterp lay on the
Neither 1 3 h7 f! xh7 ( 1 3 . . . f! a8) 14 �xh7 e4
kingside. An alternative method was 1 �e4 �e7
1 5 f! fl a3 16 �g6 a2 17 �f5 e3 18 �e4 e2 -+
2 �d4, trying to prevent the black king from join­
nor 13 �f5 f! h7 14 f!h1 f2 15 �xe5 f! xh6 -+
ing his pawn. However 2 . . . f!dS+! was strong then.
can help White.
1 hg+?
13 ... a3 14 § xe5+ ®c4 15 §e1 a2 16 h7
•••

Levenfish lets the white king go ahead for

§aS!
no reason whatsoever. An easy win was l . . . �e7
The pawns are separated by 4 files; there­
2 gh gh, for example: 3 �e4 �d7 4 f5 �c6 5 f6
fore 16 . . . f! xh7? 17 �xh7 f2 18 f! fl �d3 1 9 f! a 1 !
�b5 6 f! a 1 a4 7 �f4 f! g8 -+ , or 3 f5 ef 4 �f4
�c3 20 f! f1 ! enabled White to reach a draw.
�e6 5 �g5 �xe5 6 �xh5 �f6! with an inevi­
1 7 ®g7 f2 1 8 § f l ( 1 8 f! a 1 � b 3 )
table mate.
18 . . .a 1 � + 19 § xa1 § x a1 2 0 h8� §gl+
2 <itl xg4 <it'e7 3 ®g5 §a7 0 4 <it'h6 ®d7
White resigned.
5 <it'g7 <itlc6

1 89
 Dvoretsky 's Endgame Manual

Exercises Plan 1

It is sometimes possible to sacrifice the passed

pawn, in order to exchange rooks by means ofa
7th-rank check, transposing into a won pawn
9-125 endgame.

Benko - Gereben
9/1 9 Budapest 1 95 1
B?

9-12 7

9/20
B? If the pawn stood on a6, then after 1 �b6
the king could escape the checks at a 7. But here,
the king has no shelter, so White 's only hope lies
What is the outcome with correct play? in the exchange of rooks.
1 �b6 .§.bl + 2 �c6 .§.cl+ 3 �d6 .§.al?
The key question in a pawn endgame will
The Rook in Front of the Pawn, be: Who controls the opposition? After this mis­
with the Pawn on the 7th Rank take, it turns out to be White. Black had to con­
tinue 3 . . . E: d l +! 4 �e6 E: a l 5 E: d8 (5 E: e8 E:a6+!
We have seen a section with an identical 6 �f5 l''l xa7=) 5 . . . E: xa7 6 E: d7+ E: xd7 7 �xd7
title in the theory of "a rook and a rook pawn �h7 ! = . Note that Black must have the distant
versus a rook" endgames. The ideas from that opposition, not the close: 7 . . . �f7? 8 �d6 +- . We
section will be useful for our current consider­ examined very nearly the same situation in the
ations. pawn endings chapter (Neishtadt's study, dia­
A pawn advance to the 7th rank absolutely
gram 1 -8).
chains the opponent's forces. However, if there 4 .§.c8! (of course not 4 E: e8? E: a6+ ! )
are no vulnerable points in his camp, the game 4 . .§.a6+ 5 .§.c6 .§. x a7 6 .§.c7+ .§. xc7 7 �xc7
..

is still drawn because a king march to the pawn Now White has the di stant opposition !
is useless: no refugefrom rook checksfrom be­ There followed : 7 ... �h7 8 �d7! �g6 9 �e6
hind is provided. �g7 10 �e7 �g6 11 �f8 Black resigned.
Pushing the pawn to the 7th rank makes
Plan 2
sense, and offers winning chances, when one of
the following three plans is possible:
Sometimes the passed pawn can be ex­
changedfor some of the enemy pawns, leading
to a winning endgame with the pawns all on the
same side.
The following endgame is very important:
we shall find ourselves referring to it again and
again.

1 90
 Rook Endgames

U nzicker - Lundin Plan 3

Amsterdam ol l 954
The most important method ofplayingfor
the win with the pawn on the 7th rank is to try to
win the rook for the passed pawn. For this to
9-128 work, the enemy king must be decoyed into the
path ofa rook check (as, for instance, we tried to
do in the Unzicker-Lundin endgame, in the 4.g4+?
variation). Most often, the stronger side will try
W? to create a kingside passed pawn, which will
knock the king out of his safe square g7. A n
important point to remember is that this end
can be achieved by advancing a bishop pawn,
1 f3+! (1 a7 l"!a2+ �::,. 2 . . . �f3) 1 ... 1it>f5 2 but a knight 's or rook 's pawn is generally use­
a7! §a2+ less.
2 . . . a a6 changes nothing: 3 �d3 a d6+ 4
�c4 l"! d7 5 �c5 l"! e7 6 �d6! l"! e6+ (6 . . . l"! b7 7
l"! bS! l"! xa7 S l"! b 5 * ) 7 �d7 l"! a6 S �e7.
3 1it>d3 §a1 4 1it>d4 9-129
Observe the following tactical trick: 4 g4+
hg 5 fg+ �xg4 6 h5! . H owever in this particular
position it fails because 4 g4+? can be met with
4 . . . �f4!. w
4 §a5 5 1it>c4 §a3 6 1it>c5
••.

When B lack 's pawn stands on t7, his king

can return to f6 or g7 with an absolutely drawn
position. Here, however, White has a clear plan :
1 a7!
a king transfer to h6 fol lowed by an exchange of
This renders the f4-pawn untouchable; now
the a7-pawn for B lack's kingside pawns. B lack
the king goes after the B lack pawn, which must
has nothing to oppose this plan.
fall, because of zugzwang.
In case of6 . . . l"! xf3 7 l"!fS l"!a3 S aS� l"! xaS 9
1 1it>h7 2 1it>d3 1it>g7 3 1it>c3 1it>h7 4 1it>b3
l"! xaS 'it>g4, the simplest solution is 1 0 l"! a 3 g5
.••

§a1 5 1it>b4 1it>g7 6 1it>c5 §a6 7 1it>d5 §a1 8

1 1 hg fg 12 'it>d4 h4 13 gh gh 14 'it>e3 �g3 1 5
lit>e5! §a5+ 9 1it>e6 0 lit>h7 10 lit>f6 0 +-
a as.
It should be noted that the inexact S �e6?!
6 §a1 7 1it>d6 §a3?1 (7 . . . l"!a6+) S lit>e7?1
•••

l"!a5 leaves White, not Black, in zugzwang (9

White follows his plan, missing an immedi­
a dS?? l"! xa7 10 l"! d7+ leads to a drawn pawn
ate winning opportunity: S l"! cS! /::,. 9 l"!c5 * .
endgame). But he can easily give his opponent
s §a6
...

the move by playing 9 �d6 �h7, and now either

S . . . l"!a2 is slightly more clever; the point is 9
10 �e7 �g7 ( 1 0 . . . a a6 1 1 �f7 0 ) 1 1 �e6, or 1 0
�f7?! a a6! 1 0 �g7 g5 1 1 hg 'it>xg5 1 2 'it>t7 'it>f5
�c6 (threatening 1 1 �b6) 10 . . . l"! a 1 1 1 �d5 l"!a5+
1 3 g4+? hg 1 4 fg+ �f4, with an important cir­
12 �d6! �g7 13 �e6 0 +- .
cumstance: the f6-pawn is protected by the rook.
The squares a6 and t7 are corresponding; the
simplest way for White to circumvent the mined
square is by playing 9 �fS! a a6 1 0 �f7! 0 l"!a3
11 �g7, and 1 l . . .g5 is absolutely hopeless here.
9 1it>f7 §a3 10 lit>g7 §a1 1 l lit>h6! §a6
12 §bS §xa7 13 §b5+ lit>e6 14 1it>xg6 §a8 15
lit>xh5 §g8 16 g4 §hS+ 17 1it>g6 Black resigned.

191
 Dvoretsky 's Endgame Manual

Rovner - Shchipunov l .§a4+1 2 \t>e3 .§a3+ 3 \t>f2 .§a2+ 4

•••

Kiev 1 93 8 (sides reversed) \t>gl .§al +I (the simplest, although 4 . . . §a3 5 g4

hg 6 f4! �h7! also holds) 5 \t>g2 .§a2+ 6 \t>h3
.§a3
6 . . . § a4 7 f4 §a3 is also good.
9- 130 7 f4 .§a2 8 g4 .§a3+
The king can only escape from the checks
by approaching the rook, but this is too danger­
ous : Black takes on g4 and his g-pawn rushes to
W? the promotion square.

Let us examine two considerably more com­

plex and eventful examples.
l b7? M. Dvoretsky, 2003
As we have j ust seen Black inevitably loses
his f-pawn. B ut, in contrast to the previous ex­
ample, White fails to get a passed f-pawn. The
winning way was 1 �d4! § xf2 2 § cS §b2 3 § c7+
9-132
�g 6 4 b7 etc.
$
1 . . \t>g7 2 \t>d4 .§b5 3 \t>c4 .§b2 4 \t>d5
.

.§d2+! 5 \t>e5 .§e2+ 6 \t> x f5 .§ x f2+ 7 \t> x g4

.§b2
W?
Even two extra pawns cannot bring the ad­
vantage home. The game was drawn.

A slightly more complicated example of the

same theme . l a71
White 's plan is clear: his king will go after
R. Kholmov, 1983 the f5-pawn. Ifthe Black h-pawn were at h5, White
would win without the slightest difficulty - just
as he does in diagram 9- 1 29 . However, with the
pawn at h6 instead, Black has counterchances
9-131 involving the attempt to zip the king up in the
stalemate haven at h5 . With this configuration,
this stalemate defense is well known from pawn
endgames; it does not appear to have been em­
w ployed before in a rook endgame.
After 1 �d4? § xh2 2 'it>e5 Black saves him­
selfby playing for stalemate: 2 . . . 'it>g6! 3 a7 § a2 4
§ gS+ �h5 5 aS� § a5(e2)+ - the rook has be­
H e re agai n , two extra pawns are n o t come a desperado, or 3 §gS+ �h5 4 §fS §e2+! 5
sufficient for a w i n . B lack can easi ly prevent �xf5 (5 �d6 § a2=) 5 . . . § e8!=.
creation of a passed f-pawn. l .§a4 2 \t>d3 \t>h7 3 \t>c3
•.•

1 \t>f4
The threat is 2 g4 § a4+ (2 . . . hg 3 �xg4 b..
h4-h5 +- ) 4 �g3 hg 5 f4! , followed by h4-h5.
The immediate 1 g4 hg 2 fg leads to an obvious
draw. Black need only remember to harass his
opponent with checks from the rear when the
white king comes to the 6th rank, otherwise a
winning pawn endgame can arise.

1 92
 Rook Endgames

his rook: 10 . . . .§ e8!=. So White must give up his

a-pawn: 10 .§ d8 .§ xa7, but here I don't see a clear
way to win. For example: 1 1 'it'e6 'it'g6 1 2 f5+?
9-133
'it'g5 13 .§ g8+ 'it'h5= or 1 1 .§ d5 .§ a6.
5 ... §a5 6 �b4 §at 7 �c5 §a6 8 �d5
§a1 9 �e5 §a5+ 10 �e6 0 �g6!?
On 10 . . . <it'h7 the simplest reply is 1 1 <it>f6 0 .
B
Also possible i s 1 1 .§ d8 .§ xa7 1 2 'it'xf5, obtain­
ing a won ending with all the pawns on the same
side - though not, of course, 12 .§ d7+? .§ xd7 1 3
<it'xd7 'it'g6 1 4 'it'e6 'it'h5 !=.
3 ... �g7 11 h311 §a6+ 12 �e7 gh 13 §g8+ �h5
A clever attempt 3 . . .'�g6!?, suggested by 14 aS� §xa8 15 §xa8 �g4 16 �f6 + - .
Gansauer, is refuted by means of 4 h 3 ! ! gh
(4 . . . <it'h5 5 hg+ 'it'xg4 6 l"l g8+ 'it'h5 7 g4+!) 5
.§g8+ 'it'h5 6 a8'li¥ .§ xa8 7 .§ xa8 'it'g4 8 'it'd4 h2 9 Zurakhov - Vaisman
.§a1 'it'xg3 1 0 'it'e5 +- . Note that White wins only USSR 1 966
because his king can get to e5 in time. If White 's
king had been cut off on the 2nd rank in the start­
ing position, then as soon as it gets to c2 (or
even d2), Black plays . . . 'it'g6!, and the move h2- 9-135
h3 no longer works.
But what would happen ifBlack's king could
reach h5? We shall see about this in the next
annotation. w
4 �b3 §a6 5 �c41
5 'it'b4? is erroneous because of 5 . . . <it>g6! as
the move 6 h3 no longer works since the white
king is too far away from the f5-pawn. Interesting l a7?1
lines arise after 6 'it'c5 'it'h5! 7 'it'd5 l"la2 8 'it'e5 . Leaving the pawn on a6 made more sense.
White could have played h2-h4-h5 and l"l a7, af­
ter which his king goes to the queenside at the
cost of the g-pawn (or the f3 -pawn if he could
not avoid g2-g4). A win was rather easy because
9-134
the black king was forever locked on h7.
With his actual move, White plans h2-h4
and g2-g4-g5. After a forced double capture on
g5, he wants to take the g5 and e5 pawns with his
B?
king and to play f5-f6 thereafter.
l ... §a2
If l . . .l"l a3!? ( .6. 2 . . . e4! =) then both 2 'it'g4
and 2 'it'f2 are good: Black only postpones an
The o b v i o u s 8 . . . l"l a 5 + ? 9 'it' e 6 .§ a6 + advance of the white kingside pawns for a while
( 9 . . . <it>g6 1 0 h3! ! ) 10 'it'xf5 .§ xa7 would lose to but cannot prevent it.
1 1 h3!! .§f7+ (checkmate was threatened) 1 2 <it>e6 2 h4 §a3
l"lb7 13 hg+ ( 1 3 .§aS+ 'it'g6 14 hg l"lb3 is less 2 . . . h5? loses immediately: after 3 'it'h3 Black
convincing) 13 . . . 'it'xg4 14 l"lg8+ 'it'h5 1 5 g4+ 'it'h4 has no defense against g2-g4-g5-g6+ .
1 6 f5 . 3 �h2
8 . . . l"le2+! 9 <it'xf5 l"l e7!! (but not 9 . . . .§ e8? 1 0 This is correct: White's task is simpler when
h 3 ! ! +- ) i s m u c h stronger. I t ' s reciprocal the king hides behind the pawns. 3 'it'g4 .§ a4+ 4
zugzwang! If 1 0 'it'f6? Black can already sacrifice 'it'h5 is also playable, but then White has to show

1 93
 Dvoretsky 's Endgame Manual

more attention and accuracy. Black responded On 33 . . . .§ f6+? White could have passed the
with 4 . . . .§ a5!? and in case of the indecisive 5 g3? move to the adversary by means of triangula­
saves the game by means of 5 . . . e4! 6 fe g6+ 7 tion: 34 �e8! .§ a6 35 �e7 0 +- .
�g4 gf+ (rather than 7 . . . h5+? 8 �f4 gf 9 .§f8! 34 ®f8 (34 f6 gf 35 f5 .§ a6 0 ) 34 §.a5 •••

.§ xa7 1 0 �xf5 .§g7 11 .§ xf6 .§ xg3 12 e5) 8 ef (34 . . . .§ a6? 35 �e7 0 +- ) 35 f6 (he has nothing
�g7 9 �f4 h5=. else) 35 ...gf 36 f5 §.a1 (36 . . . .§a6? 37 �f7! 0 +- )
After 5 g4!, however, both 5 ... g6+? 6 fg+
�g7 7 g5 hg 8 hg fg (8 . . . e4 9 .§ e8!) 9 �g4! 0 +­
and 5 . . . e4? 6 g5 (6 fe?? g6 * ) 6 . . . hg 7 hg g6+ 8
fg+ �g7 9 .§ e8! +- (or 9 f4!? f5 1 0 .§ e8 .§ xa7 1 1 9-13 7
.§ e5 +- ) fail. Black has to play 5 . . . .§ a 1 6 g5 hg 7
hg fg, transposing to the game continuation.
3 ...§.a2 4 ®h3 E!.a3 5 g3 0 §.a2 6 g4 E!.a3
7 ®g2 §.a2+ 8 ®g3 §.a4 9 g5 hg 10 hg fg 1 1 W?
®h3 §.h4+ 12 ®g2 §.a4 13 ®g3 0 §.a1 14
®g4 §.g1+ 1 5 ®h5 §.a1 16 ® xg5 §.a6 17
®g4 E!.a3 18 ®h4 0 §.a4+ 19 ®h3 §.a2 20
®g3 0 E!.a5 21 ®{2 E!.a3 22 ®e2 E!.a5 23 ®d3
§.a4 24 ®c3 §.a1 25 ®b4 §.a2 26 ®c5 §.a1 37 §.e8?
27 ®d6 E!.a5 28 ®e6 §.a1 29 ®xe5 §.a6 30 f4 This natural move (White intends a transi­
E!.a5+ 31 ®e6 tion to a winning pawn endgame) is wrong. Black
has a defense based upon stalemate ! Dolmatov
suggested the correct procedure:
37 �f7! (it is important to drive the rook to
9-136 a6) 37 . . . .§ a6 38 �e7! (by the way, after the imme­
diate 37 �e7? �g7 there is no win anymore)
38 . . . �g7 (White could of course have had this
position earli er) 39 �d8! (39 �e8?! �g8!)
B? 39 ... .§ a 1 (both 39 ... �g8 40 �c7+ �g7 4 1 �b7
and 39 . . . �h6 40 �c7 �g5 41 .§ g8+ �xf5 42
a8i£t .§ xa8 43 .§ xa8 are hopeless) 40 .§ c8! .§ xa7
41 .§ c7+ .§ xc7 42 �xc7 �h6 43 �d7 �h5 44
31 ... §.a1! �e7! �g5 45 �e6 0 +- .
3 l . . . .§ a6+? loses to 32 �e7! 0 .§a4 33 f6! gf 37 §. x a7 38 §.e7+ ®h8! 39 ®f7
.•.

34 �f7! 0 .§a6 (34 . . .f5 35 �e6 .§ a5 36 �f6 0 +- ) White cannot take the rook because of stale­
35 f5 0 . The same zugzwang position (with Black mate. Hence he goes for the f6-pawn.
on move) arises after 3 l . . . .§ a4? 32 �f7! .§ a6 33 39 §.a6 (39 . . . .§a1 is also good) 40 ®g6
•••

�e7. §.aS 41 ®xf6 ®g8??

Now 32 f6 does not bring an easy win in A serious mistake when the goal was within
view of 32 . . . .§a6+ 33 �e7 (33 �f7 .§ xf6+) 33 . . . gf reach. The draw could be achieved by means of
34 �f7 (34 f5 �g7) 34 . . . .§ a4! 35 f5 .§ a6 0 . 4 l . . . .§ a6+ (4 l . . . .§al) 42 .§ e6 (42 �f7 �h7 43 f6
32 ®e7 §.a6! .§ aS, or 43 . . . .§ a 1 , but not 43 . . . .§ b6??) 42 . . . .§ a 1 !
Only this prevents the menacing advance 4 3 �g6 .§ g 1 + 4 4 � f7 �h7 (4 4 . . . .§ g7+) 4 5 f6
f5-f6. 32 . . . .§ a4? 33 f6 gf 34 �f7 0 is an error. We .§g7+! 46 �e8 (46 fg - stalemate) 46 . . . .§ g8+, and
may come to the conclusion that the squares f7- the rook returns to the long side.
a4 and e7-a6 are corresponding: this is a case of 42 ®g6 Black resigned.
reciprocal zugzwang. And if Black defends him­
self correctly he does not fall into the zugzwang.
33 ®f7 §.a4!

1 94
 Rook Endgames

Ljubojevic - Gligoric
Tral!icf)medies Belgrade m (9) 1 979
Piihtz - Kosteniuk
Mainz m ( 5) 2002

9-140

9-138

W?

1 g7 <it'b7??
l . . .c4+! 2 �b4 �b7 3 'it'b5 �a7 led to a
draw. White 's king cannot stop both the rook
The main distinction between this position
and the king at the same time : after 4 �c6 the
and the very similar endgame Benko-Gereben rook is released from the burden of protecting
(diagram 9- 1 2 7) - is that here White has a pawn the pawn.
at f3, thanks to which every possible pawn 2 c4! E!g2 3 <it'c3
endgame is won. An elementary path to victory Black resigned. The white king goes through
lay in 1 § d8! §b6+ 2 § d6 l"i: xb7 3 l"i: d7+ l"i: xd7 4 the center to the c5 -pawn and gains it by means
�xd7. of a zugzwang.
1 §eS?? E!b6+! 2 <it'f5 E! x b7 Milic and Bozic annotated this endgame for
The position has now become drawn - but the Chess Informant, Vol. 2 7. In their opinion,
the adventures have not ended yet. White could have won it after 1 c4 l"i:g3+ 2 �c2.
3 e5 fe 4 E! xe5 E!f7+ 5 <it' xg5 E! xf3 6 But they are obviously wrong : 2 . . . �d6 3 g7
E!e7+ <it'f8 7 E!a7 'it'e6(e7) leads to a drawn pawn endgame, while
after 2 �a4 �b6 3 g7 �b7 4 'it'b5 l"i: g5 5 �a4
l"i: g3! White 's king cannot break loose.
Y. Averbakh
9-139

9-141
B?

B
7 §c3??
•..

7 . . . �g8! 8 �g6 § f8= was necessary - as

we know, against a knight pawn, passive defense This example is taken from Averbakh ' s
by the rook on the 8th rank guarantees a draw. endgame handbook. It's amusing not only i n itself,
8 <it'h5?? After 8 �g6! White must reach b u t a l s o b e c au s e o f s everal grave e rrors
the "Lucena Position," for exampl e : 8 . . . § c6+ 9 c o m m itted by this fam o u s c o n n o i s seur of
�h7 l"i:c5 1 0 l"i: g7. endgame theory. Averbakh believes the position
8 <it'g8 9 E!d7 E!c6 And now we have
••• is drawn on account of l . . . l"i:a2 2 �xf5 § xf2+ 3
reached "Philidor 's Position." The game was �xg4 l"i:a2=. Black applied this defensive method
eventually drawn. in a similar situation in Rovner- Shchipunov
(diagram 9- 1 3 0) . But there, first of all, White 's
king was less active and he was unable to force
the exchange of rooks; and secondly, White 's

1 95
 Dvoretsky 's Endgame Manual

pawn was at h2, which means the pawn endgame

would still have been a draw. But here, with the
pawn at h4, the pawn endgame is won !
2 :8 e8 ! (in stead of 2 � x f5?) 2 . . . :8 x a7 9-143
(2 . . . :8 a6+ 3 �xf5 :8 xa7 4 �xg4) 3 :8 e7+ :8 xe7+
4 �xe7 �g6 5 �e6 �h5 (Black's last hope is a
chance for a stalemate) 6 �f6! 0 f4 7 gf �xh4 8
�g6 +- (or 8 f5 +- ). 9/22
Another try is l . h 5 ! ? . Levenfish and
. .
W?
Smyslov analyze this in their book on rook-and­
pawn endings. They convincingly prove that the
outcome depends on whose turn in is to move.
Black, if on move, achieves a draw as fol­ The Rook in Front of the Pawn,
lows: 2 . . . :8 a6+! 3 �xf5 :8a5+ 4 �f4 :8 a4+, or 3 with the Pawn on the 6th Rank
�e5 :8a2! (3 . . . :8 a3? is wrong in view of 4 �f4
:8a5 5 :8b8 :8 xa7 6 �g5, but 3 . . . :8a4!= is also
If a pawn advance to a 7 makes no sense,
playable) 4 �f4 :8 xf2+ 5 �g5 :8a2 6 �xh5 f4!=.
White leaves the pawn on a6 and brings his
IfWhite is on move, he wins by 2 :8 e8! :8a6+
king to the queenside where it has a refuge
(2 . . . :8 xa7 3 :8e7+) 3 �xf5 :8 xa7 4 :8e5! (Averbakh
against vertical checks. But it is a long way to
only examines 4 �g5? :8a5+ 5 �f4 :8a2=) 4 . . . �h6
go, leaving the black rook enough time to cap­
(otherwise 5 �g5) 5 :8 e6+ �g7 6 :8g6+ �h7 7
ture one or two pawns, before it must be sacri­
:8 f6! ( .l:>. 8 �g5) 7 . . . :8a5+ (7 . . . :8 g7 8 :§ f8 0 ) 8
ficed for the a-pawn. This leads to a sharp
�f4 :8a2 (8 . . . �g7 9 :Bf5) 9 �g5 :8a5+ 1 0 :8f5 +- .
Averbakh 's evaluations are the opposite : "Rook vs. Pawns " endgame, the outcome of
he suggests passing the move to the adversary. which will depend on whether White 's king can
Therefore almost all his analysis is erroneous ! get back to the kingside in time.
2 ®d6? (! Averbakh) 2 ®h7? (2 . . . :8 a6+!)
••.
For many years, it was believed that with
3 ®e7? (3 �c6! :8a2 4 :8 d8! :8 xa7 5 :8 d7+ :8 xd7 correct defense, the draw was an easy matter,
6 �xd7 and 3 �e6! �g7 4 :8 e8! or 3 . . . :8 a6+ 4 something B lack could achieve with a couple of
�e5! �g7 5 �f4! are winning) 3 ®g7 4 ®e6 ...
tempi to spare. This point of view was espoused
§a2? (4 . . . :8 a6+!) 5 ® xf5? (5 :8e8! +- ) 5 §a5+ •.•
in, among other places, the first German editions
(5 . . . §. xf2+? 6 �g5 :9.a2 7 �xh5 §. a4 8 :9. e8 §. xa7 of this Manual.
9 �xg4) 6 ®f4 ®h7?? (6 . . . :9. a4+ is a draw) 7 But in the latter half of 2003 , the theory of
§f8! § x a7 8 ®g5 §a5+ 9 §f5 +- . this portion of the endgame underwent some revo­
lutionary changes. Black's position, it turned out,
was far more dangerous than it had seemed.
Exercises Johannes Steckner, a Swiss player, while
checking the analysis of one of the basic posi­
These two exercises are not complicated; in tions which had been considered drawn, found a
fact, they could have been included in the previ­ tremendous improvement for White, leading to a
ous "tragicomedies." win for him. And his discovery led, in turn, to
new researches that were conducted by Steckner,
grandmasters Karsten M u l l e r and Rustem
Dautov, and myself; along with other endgame
aficionados that came upon our researches in
9-142 the chess press.
Here I shall present only the most important
analyses. For those who seek more detailed in­
fo rmat i o n , I w o u l d r e c o m m e n d v i s i t i n g
9/2 1 www.chesscafe.com and looking for m y articles
B? entitled, Theoretical Discoveries, as well as vari­
ous articles authored by Karsten Muller.

1 96
 Rook Endgames

Nevertheless, even in edited form, the ma­ more logical approach is to begin counterplay
terial I offer for your consideration is so large immediately by 2 .. .f6!? (2 . . . �g4 3 .§ xf7 .§ xa6
and complex, that it clearly exceeds the bound­ doesn't lose, either). After 3 �f3 g5! 4 hg fg 5
aries I tried to maintain when I wrote this Manual. .§ aS g4+ 6 <;£;le3 <;£;>g6 the king gets back to g7, so
The excuse I offer is its newness and enormous White plays 3 .§ aS instead, threatening to obtain
practical significance to the theory of this sort of a winning position with the pawn on the 7th,
endgame. known to us from the Unzicker - Lundin game
Nothing could be further from my mind than (diagram 9- 1 28), by 4 f3 .§ a3+ 5 �e2 .§a2+ 6
to label the analysis presented below as the "last �d1 .§a3 7 a7. But Black draws by playing
word of theory" - long, complicated variations 3 . . . <;£7g4 4 a7 f5 !? (4 . . . .§ a3+ 5 <;£;le4 f5 + 6 <;£;le5
rarely turn out error-free. But in any case, they �f3= is good too) 5 .§ gS f4+! 6 gf .§a3+ 7 <;£;le4
go a long way to correct and develop the pre­ (7 �e2 .§ xa7 S .§ xg6+ <;£;>xf4=) 7 . . . .§ a4+ S <;£;le5
existing conclusions, and may in turn serve as a l"la5+ 9 �e6 .§a6+ 10 <;£;>f7 .§ xa7+ 1 1 <;£;lxg6 .§a6+
starting point for additional theoretical re­ 12 �f7+ �xf4= (analysis by Dvoretsky) .
searches. l §. x f2 2 E!c7! §.a2 3 a7
...

On 3 .§ c6+? <;£7f5 4 <;£;lc5 <;£7g4 5 <;£7b5 <;£;lxg3 6

V. Kantorovich, 1988 .§ c4 f6! 7 .§ a4 .§ b 2 + S <;£;>c6 .§ bS 9 a7 .§ aS
J. Steckner, 2003 (Kantorovich), Black does indeed obtain a draw
with two tempi to spare.
3 <i!;f5
...

Kantorovich's analysis continued 4 .§ xf7+

<;£7g4 5 <;£;lc5 <;£;lxg3 6 �b5 ! .§b2+! 7 <;£;>c6 .§a2 S
<;£;lb7 <;£;lxh4 9 .§ f6 .§ xa7+ - here too, the draw is
completely obvious.
It was Steckner who offered the powerful
improvement: 4 <i!;c4!!.

9-145
In 1 989, Vadim Kantorovich, of Moscow,
published an interesting article titled, The Out­
side Passed Pawn . The article opened with the
diagrammed position. The main conclusion of the
B
analysis was: Black draws with two tempi to spare.
But in fact, he's lost !
1 <i!;d4!
The pawn must be sacrificed precisely with
the rook on a7 ! 1 .§ aS? <;£7f5 would be much H i s idea becomes c lear in the variation
weaker. Please note that Black �\' pieces are opti­ 4 ... <i!;g4 5 ®b3! §.a6 6 §.c4+ <i!; xg3 7 §.a4.
mally placed: the rook holds the f-pawn in the Now White forces the sacrifice of B lack's
crosshairs, while the king occupies the most rook without wasting time on the king's long
active available square. march to a7 , and wins move-on-move ("Chess is
Black would have an easy draw after 2 <;£7d4 the tragedy of a single tempo ! ") .
.§ xf2 3 .§ fS .§a2 4 .§ xf7+ <;£;>g4= or 2 f3 .§ a3+ 3 7 ... §. x a7 8 §. xa7
<;£7d4 .§ xf3 4 .§fS .§a3 5 .§ xf7+ <;£7g4 6 .§f6 <;£;lxg3 7
.§ xg6+ <;£;>xh4 S <;£;lc5 <;£7h3 9 <;£7b6 h4= (both lines
by Kopaev).
On 2 .§a7!? retreating the king by 2 . . . <;£;>f6? or
2 . . . �e6? loses, as will become clear later on. A

1 97
 Dvoretsky 's Endgame Manual

Alas, White wins here too, with 9 l"l g7!.

9-146

9-148

8 ... \t' x h4 9 \t'c3 (there 's no time to take

the pawn: 9 l"l xf7? �g3=) 9 ... \t'g3 a) 9 . . . g5 1 0 �b8 �g4 1 1 a8� (but not 1 1
9 . . .f5 10 �d3 g5 doesn' t help : the rook can hg? fg 1 2 a8� l"l xa8+ 1 3 �xa8 h4 1 4 gh �xh4
deal with all three pawns. 1 5 �b7 g4 1 6 �c6 � g 3 ! 1 7 � d 5 �f3 =)
10 \t'd2 h4 ( 1 0 . . . g5 11 l"l xf7 +- ) 1 1 \t'e2
1 1 . . . /"l xa8+ 1 2 �xa8 �xg3 1 3 hg fg 1 4 l"l xg5+
\t'g2 (l l . . .h3 12 �fl) 12 E!, x f7 h3 13 E!,f2+1
�h4 15 l"l g8 �h3 16 �b7, and the king gets
\t'g3 14 E!.f6 +-
back to f3 in time.
Let's try a different defense, such as 4 . . . 1"l a 1 ,
b) 9 . . . �g4 1 0 /"l xg6+ �h3 1 1 l"l g7 l"la3 1 2
getting the rook away from the tempo-gaining
�b8 l"lb3+ 1 3 l"lb7 l"l xg3 1 4 �c7!? ( 1 4 l"lb4 i s
�b3. After 5 �b5 however, the only way to fore­
also strong) 1 4 . . . 1"la3 (14 . . . 1"l g8 1 5 l"lb8 l"lg7+ 1 6
stall the threat of closing off the a-file is by a
�b6 l"l xa7 1 7 �xa7 +- ) 1 5 l"lb3+! l"l xb3 16 a8�.
series of checks, which drive the white king for­
At the start of this section, I presented a
ward: 5 . . . 1"lb1 + 6 �c6 l"l a 1 7 �b7 l"lb1 + 8 �c8
formula in the most general terms for how play
l"l a 1 9 l"l xf7+ �g4 .
might develop in this kind of ending. Now we
can more precisely restate White 's most danger­
ous plan. The pawn advances to a6; the rook
stands on a 7, and at the first opportunity will
9-1 4 7
move aside to c 7, clearing the path ofthe pawn.
White 's king selects a path to advance which
will allow him to execute the interference idea
as quickly as possible - that is, moving the rook
W?
with tempo to the a-file.

Let's return to the starting position for this

endgame - diagram 9- 1 44 and ask ourselves
Here 1 0 �b7 1"1 b 1 + is useless; and 1 0 �b8?
-

�xg3 1 1 l"lf6 �xh4 1 2 /"l xg6 �h3 13 �b7 /"l xa7+ this question : can B lack save himself if he is on
is only a draw. White gains the tempo he needs the move? And the draw turns out to be no simple
by 10 l"lg7! �xg3 11 /"l xg6+ �xh4 1 2 �b7 l"l xa7+ thing to achieve in this case, either.
( 1 3 l"l a6 was threatened) 13 �xa7 �h3 14 �b6 I. I began my testing with the obvious move
h4 1 5 �c5 �h2 16 �d4 h3 17 �e3 �h1 1 8 1 ... \t'e5, and quickly found the line 2 \t'd3!
�f3 +- . E!. xf2 3 E!.e7+1 (the immediate 3 l"l c7 also de­
The only thing left to try is 4 . . .f6. White serves study) 3 ... \t'f6 4 a7 E!.a2 5 E!.c7 \t'f5 6
can 't respond with 5 �b4? �g4 6 �b3 l"l a6 7 \t'c41, leading to the position in diagram 9- 1 45 ,
l"l c4+ �xg3 8 l"l a4 l"l xa7 - by comparison with i n which Steckner demonstrated the win for White.
the line 4 . . . �g4, Black has gained the useful But what if Black chooses 4 . . . �xe7 5 a8�
move . . . f7-f6, which alters the assessment of the l"l f5, hoping to set up a rook-vs. -queen for­
position (9 /"l xa7 g5=). Now comes a series of tre s s ?
forced move s : 5 �b5 l"l b2+ 6 �c6 l"la2 7 �b7 Look i n Chapter 1 3 there you will find a
-

l"lb2+ 8 �c8 l"la2. similar position that occurred in the game Dorfinan

1 98
 Rook Endgames

l"\ g8!?, after which l l . . . �xg3?? i s bad: 1 2 hg fg

1 3 l"\ xg5+ and 1 4 l"\ a5, while 1 l . . .l"\b2+? 1 2 �c5
l"\a2 1 3 a8i;1t l"\ xa8 14 l"\ xa8 �xg3 1 5 �d4! �xh4
9-149
16 �e3(e4) leads to a position in which the rook
more than likely wins against the three pawns.
Black can secure the draw by means of the wait­
ing move 1 1 . . . l"\ a 1 ! , for example, 1 2 �b7 l"\b1 +
w
1 3 �c6 l"\a1 14 a8i;1t l"\ xa8 1 5 l"\ xa8 �xg3 1 6
�d5 gh=.
b) The strongest line is 7 .§g7! �c6 (after
7 . . . �e6 Black has a tempo less in comparison
- Beliavsky (diagram 1 3-33) and which shows that with the previous variation, and loses after 8 �b5
Black (or in that game - White) was quite correct �f5 9 �b6 g5 10 �b7) 8 .§f7! (but not 8 l"\ xg6?
to expect a draw, except that his king had to be on l"\ xa7 9 l"l. xf6+ �d5 10 l"\f5+ �e4 1 1 l"\ xh5 l"\g7=)
g7. With the king stuck in the center, however, he 8 f5 9 .§g7 �b6 10 �c4, when White must
...

loses. win.
In the variation we have j ust examined: 11. l �e6 2 �d4! f6 (we already know
...

l . . .�e5 2 �d3 l"\ xf2 3 l"\ e7+, B lack could keep the consequences of 2 . . . l"\ xf2 3 l"\ c7 l"\ a2 4
his king in the center: 3 . . . �d5 (or 3 . . . �d6) 4 a7 a7 +- ).
l"\a2. White continues 5 l"\ xf7 (and with the king Steckner demonstrated the win for White
on d5 , perhaps, 5 �c3!?) and wins by attacking after 3 �c5 �f5 4 f3! l"l. a3 5 �b4 l"l. xf3 .
the kingside pawns at the appropriate moment
with his rook. Let's examine a characteristic and
quite important variation. Steckner uncovered it,
while I have added a few explanations and 9-151
touched some things up.
l �e5 2 �d3 �d5 (instead of 2 . . . l"\ xf2)
...

3 �c3 .§ xf2 4 .§c7 .§a2 5 a7 f6 6 �b4 �d6

a) White only gets a draw after 7 §.f7 �e6 w
8 l"\ g7? (8 l"\ c7) 8. .. �f5 9 �b5 g5 (Black could

This is a good time to draw your attention

9-150 to a problem that must often be resolved : which
is the best square for the rook - b7 or c7 (or, with
the rook on a8 - b8 or c8)? Sometimes, the choice
is made on purely tactical considerations : for ex­
W? ample, with the king at c5 and the rook at a8, the
move l"\ c8 would be impossible because of
... l"\c3+. And if, with the rook at a7, B lack 's rook
were to occupy the 8th rank, then it would make
transpose the last two moves) 1 0 �b6 �g4. Now sense to continue l"\b7 and a6-a7, creating the
1 1 hg fg 12 �b7 h4 ( 1 2 . . . l"\b2+!?) 13 gh �xh4 threat of l"\b8. But, it seems to me that it most
14 l"\ g6 l"\ xa7+ 1 5 �xa7 g4 16 �b6 �g3! 1 7 often makes sense to retreat the rook to the c­
�c5 �f3= is harmless. If l l �b7, the immediate file. In that case White 's king on the b-file will
capture on g3 loses - first, Black must drive back not hinder the rook 's mobi lity; and the threat of
the White king: 1 l . . .l"l.b2+! 12 �c8 l"\a2 1 3 �b8; checks from the side by White 's rook followed
only now can he play 1 3 . . . �xg3 14 hg fg 1 5 by the a-file interference becomes more realistic .
l"\ xg5+ �h4= . The most dangerous try i s : 1 1 Naturally, I cannot prove my assertion; I
can only provide illustrations.

1 99