Journal of Sports Sciences

ISSN: 0264-0414 (Print) 1466-447X (Online) Journal homepage: https://www.tandfonline.com/loi/rjsp20

Testing the application of corrective adjustment

procedures for removal of relative age effects in
female youth swimming

Shaun Abbott, Kylie Moulds, James Salter, Michael Romann, Lucy Edwards &
Stephen Cobley

To cite this article: Shaun Abbott, Kylie Moulds, James Salter, Michael Romann, Lucy Edwards
& Stephen Cobley (2020): Testing the application of corrective adjustment procedures for
removal of relative age effects in female youth swimming, Journal of Sports Sciences, DOI:
10.1080/02640414.2020.1741956

To link to this article: https://doi.org/10.1080/02640414.2020.1741956

View supplementary material

Published online: 21 Mar 2020.

Submit your article to this journal

Article views: 29

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=rjsp20
JOURNAL OF SPORTS SCIENCES
https://doi.org/10.1080/02640414.2020.1741956

SPORTS PERFORMANCE

Testing the application of corrective adjustment procedures for removal of relative

age effects in female youth swimming
a
Shaun Abbott , Kylie Mouldsa, James Salterb, Michael Romann c
, Lucy Edwardsa and Stephen Cobley a

a
Discipline of Exercise & Sport Science, Faculty of Health Sciences, The University of Sydney, Sydney, Australia; bSwimming Australia Ltd, Sunnybank,
Australia; cSwiss Federal Institute of Sport, Magglingen, Switzerland

ABSTRACT ARTICLE HISTORY

The purpose of this study was (1) accurately estimate longitudinal relationships between decimal age (i.e., Accepted 20 February 2020
chronological and relative) and performance in Australian female 100 m (N = 765) and 200 m (N = 428)
KEYWORDS
Breaststroke swimmers (10–18 years); and (2) determine whether corrective adjustment procedures could Athlete development; talent
remove Relative Age Effects (RAEs) in an independent sample of age-matched 100 m (N = 2,491) and 200 identification; swimming;
m (N = 1,698) state/national level Breaststroke swimmers. In Part 1, growth curve modelling quantified youth competition
longitudinal relationships between decimal age and swimming performance. In Part 2, relative age
distributions (Quartile 1–4) for “All”, “Top 25%” and “10%” of swimming times were examined based on
raw and correctively adjusted swim times for age-groups. Based on raw swim times, finding identified
RAE effect sizes increased in magnitude (small-medium) with selection level (“All”-“Top 25%”) in 12–14
years age-groups for both events. However, when correctively adjusted swim performances were
examined, RAEs were primarily absent across all age-groups and selection levels. Using longitudinal
reference data, corrective adjustment procedures removed relative age advantages in female youth
Breaststroke performance. Removing the influence of relative age-related differences is predicted to
improve the accuracy of identifying genuinely skilled youth swimmers.

Introduction Helsen et al., 2005), Basketball (López de Subijana & Lorenzo,

2018) and Rugby (Till et al., 2010), participation ratios between
The common practice of placing children into annual age
the relatively older and younger quartiles (Odds Ratios (OR))
groups for competitive sport can unintentionally undermine
have varied from small (e.g., OR = 1.3 -1.7) to large (e.g., OR ≥
the objectives of policy-makers, sporting organisations and
6.7). While RAE effect sizes are typically lower in females com-
practitioners for long-term participation and success.
pared with males during junior and adolescence, higher mag-
Administrators typically group children according to age with
nitude RAEs are still associated with selective representative
respect to date cut-off criteria (i.e., age at day of competition/
contexts at ages and time points associated with puberty and
game) for logistical and developmental control purposes to
maturation irrespective of sex (Smith et al., 2018). With few
help ensure equal competition experiences (Cobley et al.,
exceptions (e.g., gymnastics) (Hancock et al., 2015), RAEs in
2009). However, the consequence can be Relative Age Effects
female competitive swimming, similar to other sports, increase
(RAEs); an outcome which has gained significant interest in
from the youngest age divisions through to puberty, before
both sport and education settings, due to its differential out-
decreasing into adulthood (Delorme et al., 2010; Okazaki et al.,
comes on individuals (Furley et al., 2016; Vestheim et al., 2019).
2011; Smith et al., 2018).
RAEs reflect a biased distribution of participants within a given
Several hypotheses have been proposed to account for RAEs
cohort, commonly an overrepresentation of the relatively older
(Cobley & Till, 2017; Cumming et al., 2019; Hancock et al., 2013;
within an age-group cohort, and an underrepresentation of the
Musch & Grondin, 2001). Generally speaking, the most sup-
relatively younger (Wattie et al., 2008). Being relatively older is
ported based on meta-analytical and individual study findings
associated with consistent attainment and selection advan-
is the “maturation-selection hypothesis” (Cobley et al., 2009;
tages which influences the likelihood of longer-term participa-
Helsen et al., 1998; Lovell et al., 2015). The hypothesis suggests
tion (Cobley & Till, 2017; Delorme et al., 2009; Lemez et al.,
that compared with relatively younger peers, relatively older
2016) in numerous male (Cobley et al., 2009) and female (Smith
athletes have an increased likelihood of advanced normative
et al., 2018) sporting contexts.
anthropometric development (e.g., stature, body mass) along
Across various individual and team-based youth sporting
with associated superior physical capacities such as aerobic
tiers (i.e., school, local community, representative and interna-
power, muscular strength, endurance and speed (Malina et al.,
tional) notably Track & Field (Nakata & Sakamoto, 2012;
2007; Viru et al., 1999). These characteristics are important
Romann & Cobley, 2015), Tennis (Gerdin et al., 2018),
contributory factors for performance in physically demanding
Swimming (Cobley et al., 2018), Soccer (Brustio et al., 2018;

CONTACT Stephen Cobley stephen.cobley@sydney.edu.au Discipline of Exercise & Sport Science, Faculty of Health Sciences, The University of Sydney,
Cumberland Campus, 75 East Street, Lidcombe, NSW 2141, Australia
Supplemental data for this article can be accessed here.
© 2020 Informa UK Limited, trading as Taylor & Francis Group
2 S. ABBOTT ET AL.

sports. Due to the increased inter-individual variation in the differences and thereby potentially improving the accuracy of
timing and tempo of anthropometric and physical develop- performance evaluation.
ment during maturation (Cobley et al., 2014; Till et al., 2013),
the relatively younger adolescent athletes can experience
short-term performance disadvantages (Myburgh et al., 2016; Methods
Till et al., 2014) reflected by lower likelihoods of selection to Part 1: Longitudinal relationships between decimal age
representative tiers within a sport (Brustio et al., 2018; Helsen and swim performance
et al., 2005). In the longer-term, recent studies suggest that
RAE-related and maturation inequalities may be transient and Participants
temporary (Cobley et al., 2018; Deaner, 2013), yet the immedi- Participants were N = 765 and N = 428 female swimmers, aged
ate psychological and social consequences may – to some 10–18 years, who participated in official long-course 100 m and
degree – account for lower sporting involvement in junior 200 m Breaststroke events (N = 288) respectively. All partici-
and adolescent years. pants were competing in age-group and/or open-level
To help address relative age and developmental inequalities Australian domestic competitions between 1999 and 2017
in youth sport such as participation and selection disadvan- (inclusive) and were Australian residents. Participants were
tages afforded to relatively younger athletes (Hancock et al., included if they registered a state-level qualification time in
2015; Smith et al., 2018), a range of feasible strategies for the specified events; registered multiple years (≥5 years) of
national sporting organisations and practitioners have been performance times at least once per year ranging from 10 to
proposed (Cobley, 2016). Within individual sporting context 18 years; and, who were without a disability. Such criteria
(e.g., track and field, swimming), the application of corrective helped establish an accurate longitudinal estimate of perfor-
performance adjustments could help remove performance dif- mance change over time both within and across annual-age
ferences based on relative age (Cobley et al., 2019; Romann & groups.
Cobley, 2015). Corrective adjustment procedures are applicable
where performance in sport-specific tasks can be objectively Procedure
measured (e.g., centimetres, seconds, metres per second) and Following University ethics approval (App No: 2017/650), an
where the age–performance relationship can be quantified anonymised dataset containing N = 61,458 registered female
(Romann & Cobley, 2015). Based on substantial longitudinal swims in the 100 m (N = 39,721) and 200 m (N = 21,737)
data, a cohort age-performance trendline is used to adjust the Breaststroke was provided by Swimming Australia. The 100 m
performance times of relatively younger athletes so to be more Breaststroke was sampled due to the event being one of the
equitably compared with the relatively oldest in a given chron- most competitive events (i.e., higher participation numbers) in
ological age-group (e.g., Under 12’s). Australia’s female age-group championship schedule. The
Cobley et al., (2019) recently explored the application of 200 m was also sampled as a comparative event, observing
corrective adjustments to remove RAEs in junior Australian whether distance changed the nature of the relationship
male swimmers. Using a large longitudinal reference data set between decimal age and performance over time, as well as
(N = 553) aged from 10 to 18 year-olds, expected performance the nature of corrective adjustment procedures and its conse-
differences from being 1 day to 1 year older in each annual age quences on RAEs. Youth performance in both events is consid-
group were calculated. Individual performance times were then ered informative for athlete evaluation, selection and transfer
adjusted – given their chronological age in the 100 metre (m) (i.e., taking up other strokes) purposes with Swimming
Freestyle event – to a standard reference point (i.e., the rela- Australia’s development system.
tively oldest swimmer) generating a corrected swim perfor-
mance time. The relative age distributions of corrected sprint Data-analysis
performance times were then re-examined. Findings identified Extracted data, specific to the 100 m & 200 m Breaststroke, were
that for almost all annual age-groups, relative age attainment initially screened for outliers (i.e., residuals) using box plots.
discrepancies were removed – if not at least reduced – as RAEs Outliers were removed if an input data error was apparent.
became absent in the “Top 50%”, “Top 25%” and “Top 10%” of Participants were also removed if registered swim performance
100 m Freestyle times. Consequently, these findings suggested had not reached within 1 second (s) of the 100 m or 2 s of the
merit in further testing the utilisation of corrective adjustments 200 m state-level qualifying criteria by age 16 years.
in youth sports contexts where the relatively younger are dis- A normative distribution was checked for all those identified
advantaged in age-based competition. Therefore, the present in each stroke event (i.e., N = 765–100 m; N = 428–200 m). Next,
study further tested the application of corrective adjustments swimmers’ exact decimal age (i.e., years and days old) at respec-
in swimming, by extending to female swimmers in a different tive competitive events was plotted against 100 m and 200 m
stroke and at two alternative distances. Breaststroke performance times using longitudinal growth
The purposes of the present study were first (Part 1) to curve models with application of a multi-level modelling frame-
accurately estimate the longitudinal relationship between dec- work (Dormehl et al., 2016; Vacher et al., 2017). Decimal age was
imal age (i.e., chronological and relative) and performance in centred to zero, representing the first point of observation (e.g.,
female Breaststroke swimming. The second purpose (Part 2) 10.00 years of age) and acted as the independent variable.
was to determine whether corrective adjustment procedures A hierarchical method was used where repeated observations
could remove RAEs in female swimming, by controlling for age were nested within individual swimmers. An unstructured
 JOURNAL OF SPORTS SCIENCES 3

covariance type was applied, and model fit for fixed and ran- adjusted race performance times within each annual age group
dom effects (e.g., intercept and slope) were assessed by com- (12–15 years) were re-examined.
paring the log-likelihoods (−2LL) with changes in critical values
for the chi-square statistic and degrees of freedom. The final Data-analysis
fixed effect estimate model was a quadratic function (y = ax2 To examine and compare relative age distributions to expected
+ bx + c), summarising the expected decimal age–performance normative and equal birth distributions in the Australian popu-
relationship across ages 10–18 years. The specific function for lation (Australian Bureau of Statistics, 2014) for “All” swimmers,
each event was subsequently taken forward for application to quartile distributions of “Raw Top 25%–Top 10%” and the dis-
corrective adjustment calculations. tributions of the “Correctively Adjusted Top 25%–Top 10%” at
each age-age group (12–15 years), Chi-square tests (X2) were
Part 2: Application of corrective adjustments on relative applied with p set at 0.05. Post-hoc Cramer’s V determined the
age distributions magnitude of effect size between frequency count distribu-
tions. Odds Ratios (ORs) then provided more specific relative
Participants age quartile comparisons. For df = 3, which was the case for all
Participants were an independent sample of female 100 m comparisons of relative age quartiles, 0.06 < V ≤ 0.17 indicated
Breaststroke (N = 2,491) and 200 m Breaststroke (N = 1,698) a “small effect”; 0.17 < V < 0.29 a “medium effect”; and, V ≥ 0.29
swimmers aged 12–15 years who registered performance(s) at a large effect (Cramer, 1999). ORs and matching 95%
state (N = 27) and/or national (N = 5) long-course domestic Confidence Intervals (CI) estimated effect sizes of specific com-
events from 2013 to 2018 (inclusive). For each participant, their parisons (e.g., Q1 v Q4) with Q4 acting as the referent group.
fastest annual performance was extracted from across events in
a given annual-age cohort.
Results
Procedure Part 1: Longitudinal relationships between decimal age
As outlined in Part 1, similar data collection, extraction and per- and swim performance
formance criteria were implemented, with data reflecting indivi-
dual performances at long-course state and national Figure 1(a,b) summarise the curvilinear (quadratic) longitudinal
competitions. Based on the respective swim events sampled and relationship between decimal age (i.e., including chronological
dates applied for annual-age grouping, participants were assigned and relative age) for the 100 m and 200 m Breaststroke swimming
to chronological age (e.g., 12 years old) and relative age quartiles performances examined. Estimate of the relationship included
given their decimal age. For example, at 12 years old, quartile intercept, linear and quadratic components. Decimal age
categories were Q1 = 12.75–12.99 years; Q2 = 12.50–12.74; (squared) significantly predicted performance in both the
Q3 = 12.25–12.49 years and Q4 = 12.00–12.24. Then, the frequency 100 m and 200 m distances (i.e., F(1, 608.6) = 2131.9, p < 0.001
and percentage distributions of swimmers within each age group and F(1, 324.4) = 822.6, p < 0.001) according to final fixed effects
(12–15 years old) and according to relative age quartiles (Q1–Q4) estimates. The quadratic age relationship with performance
were determined (see Table 1). These steps permitted an assess- showed significant variance in intercepts (Var(u0 j) = 1.49, X2
ment of relative age distributions for “All” swimmers sampled at (1) = 2,334, p < 0.001) and slopes for the linear and quadratic
each age group. For each age group examined (12–15 years), the component (Var(u1 j) = 0.07, X2(1) = 91.64, p < 0.001) across 100 m
relative age distributions of 100 m and 200 m Breaststroke swim- Breaststroke swimmers. In the 200 m Breaststroke, the curvilinear
mers were then sub-examined according to the “Top 25%” and relationship between age and performance also showed signifi-
“10%” of performance times. Such a procedure helped resemble cant variance in intercepts (Var(u0 j) = 9.40, X2(1) = 1,251,
the introduction of selection criteria, similar to representative p < 0.001) and slopes (Var(u1 j) = 0.38, X2(1) = 121.1, p < 0.001).
levels of competition and selection. The relative age distributions In addition, the slopes and intercepts negatively and significantly
within the “Top 25%” and “10%” were then examined, with inter- covaried (100 m Cov(u0 j, uoj) = −0.68, X2(3) = 863.73, p < 0.001;
est in determining whether RAE effect sizes changed according to 200 m Cov(u0 j, uoj) = −3.93, X2(3) = 452.2, p < 0.001, respectively).
selection level. The final fixed and random effects models showed decimal age
To test whether corrective adjustments could remove RAEs had a significant negative linear and positive quadratic relation-
across age-groups (12–15) and according to performance level ship with performance time in both distances.
(i.e., “Top 25%” and “10%”), all raw performance times were
adjusted using expected within annual-age performance differ-
Part 2: Raw distributions
ences generated from the quadratic estimates described in Part
1. Thus, individual performance times registered at a given Table 1 summarises results from analysis of “All” sampled
decimal age were adjusted using the expected longitudinal swimmers aged 12–15 year-old and according to applied selec-
trend line (identified in Part 1) to the relatively oldest decimal tion criteria for raw (unadjusted) swimming times. For “All” the
age within each age-group. For example, for two females in the sample, classical RAEs were evident in both Breaststroke events
12 years age-group, one turning 12 years old on the first day of at 12–14 years of age with small effect sizes apparent. RAEs
eligibility (i.e., 12.00) and the second who was 12.99 years on then dissipated by 15 years of age in both events. However,
the day of competition had their 100 m Breaststroke time when applying selection criteria on raw swimming times (i.e.,
reduced by −4.65 s and 0.00 s, respectively. Following correc- “Top 25%” -‘10%), RAEs extended into the 15 years age-group
tive adjustments, distributions of the “Top 25%” and “10%” of for the 100 m Breaststroke only (e.g., Top 25% – X2 = 9.17,
 4
S. ABBOTT ET AL.

Table 1. Relative age distribution, chi-square and odds ratio analysis of female 100 m and 200 m Breaststroke swimmers (aged 12–15 years) at state and national level championships for 2013–2018 (inclusive). Data for the Top
25%, 10% & 5% of swim times are based on raw 100 m and 200 m swim times.
Distance & performance level Age-Group Total N Q1% Q2% Q3% Q4% X2 P V ES cat. OR Q1vQ4 (95%CI) OR Q2vQ4 (95%CI) OR Q3vQ4 (95%CI)
All 100 m swimmers 12 years 657 31.66 25.57 25.88 16.89 29.23 0.001* 0.12 small 1.87* (1.37–2.57) 1.51* (1.10–2.09) 1.53 (1.11–2.12)
13 years 643 30.48 26.59 26.13 16.80 26.00 0.001* 0.12 small 1.81* (1.32–2.50) 1.58* (1.14–2.19) 1.56 (1.12–2.15)
14 years 635 27.56 24.72 27.56 20.16 9.30 0.03* 0.07 small 1.37* (1.00–1.88) 1.23 (0.89–1.69) 1.37 (1.00–1.88)
15 years 556 26.44 25.54 23.92 24.10 0.97 0.81 0.02 no 1.10 (0.79–1.53) 1.06 (0.76–1.48) 0.99 (0.71–1.39)
Top 25% 12 years 165 38.18 27.27 21.21 13.33 21.74 0.001* 0.21 medium 2.86* (1.50–5.48) 2.05* (1.05–3.99) 1.59 (0.80–3.16)
of 100 m swim times
13 years 161 39.13 29.81 19.88 11.18 28.34 0.001* 0.24 medium 3.50* (1.77–6.92) 2.67* (1.33–5.35) 1.78 (0.86–3.67)
14 years 159 31.45 32.70 22.64 13.21 15.61 0.001* 0.18 medium 2.38* (1.21–4.67) 2.48* (1.27–4.84) 1.71 (0.86–3.43)
15 years 139 32.37 30.21 20.14 17.27 9.17 0.03* 0.15 small 1.87* (0.95–3.71) 1.75* (0.88–3.48) 1.17 (0.57–2.40)
Top 10% 12 years 66 36.36 25.76 27.27 10.61 9.02 0.03* 0.21 medium 3.43* (1.16–10.13) 2.43* (0.80–7.39) 2.57 (0.85–7.78)
of 100 m swim times
13 years 65 38.46 27.69 20.00 13.85 8.78 0.03* 0.21 medium 2.78* (1.00–7.75) 2.00* (0.70–5.74) 1.44 (0.48–4.31)
14 years 64 25.00 39.10 26.56 9.34 11.43 0.01* 0.24 medium 2.68* (0.83–8.60) 4.19* (1.35–12.96) 2.84 (0.89–9.08)
15 years 56 35.71 25.00 26.79 12.50 6.14 0.10 0.19 medium 2.86 (0.92–8.89) 2.00 (0.62–6.45) 2.14 (0.67–6.86)
All 200 m swimmers 12 years 444 30.63 26.35 25.68 17.34 16.46 0.001* 0.11 small 1.77* (1.20–2.59) 1.52* (1.03–2.24) 1.48 (1.00–2.19)
13 years 432 34.49 24.54 25.93 15.04 32.89 0.001* 0.16 small 2.29* (1.54–3.40) 1.63* (1.08–2.45) 1.72 (1.15–2.59)
14 years 436 27.75 25.92 26.83 19.50 7.33 0.06 0.07 small 1.42 (0.97–2.09) 1.33 (0.90–1.96) 1.38 (0.94–2.02)
15 years 386 27.20 24.35 23.58 24.87 1.13 0.77 0.03 no 1.09 (0.74–1.62) 0.98 (0.66–1.46) 0.95 (0.63–1.42)
Top 25% of 200 m swim times 12 years 111 37.84 30.63 22.52 9.01 20.35 0.001* 0.25 medium 4.20* (1.76–10.00) 3.40* (1.41–8.19) 2.50 (1.01–6.16)
13 years 108 44.44 29.63 18.52 7.41 32.43 0.001* 0.32 large 6.00* (2.39–15.03) 4.00* (1.56–10.24) 2.50 (0.94–6.65)
14 years 109 30.28 28.44 27.52 13.76 7.52 0.06 0.15 small 2.20 (0.98–4.94) 2.07 (0.92–4.66) 2.00 (0.88–4.52)
15 years 97 25.77 28.87 19.59 25.77 1.76 0.62 0.08 small 1.00 (0.45–2.20) 1.12 (0.51–2.44) 0.76 (0.33–1.73)
Top 10% 12 years 45 35.56 31.11 20.00 13.33 5.58 0.13 0.20 medium 2.67 (0.76–9.31) 2.33 (0.66–8.26) 1.50 (0.40–5.63)
of 200 m swim times
13 years 44 50.00 25.00 25.00 0.50 20.78 0.001* 0.40 large 44.00* (2.36–8.20) 22.00* (1.15–422.36) 22.00* (1.15–422.36)
14 years 44 36.36 27.27 22.72 13.65 4.72 0.19 0.19 medium 2.66 (0.76–9.36) 2.00 (0.55–7.24) 1.66 (0.45–6.18)
15 years 39 28.21 33.33 15.39 23.07 2.74 0.43 0.15 small 1.22 (0.35–4.27) 1.44 (0.42–4.94) 0.67 (0.17–2.60)
Q1 to Q4 = Quartile 1 to 4; Q1–Q4% = Relative age quartile (3 months combined) percentage of total number; χ2 = Chi-Square value; P = Probability value; * = Significance p < 0.05; V = Cramer’s V effect size. ES cat. = Effect Size
category; OR = Odds Ratio; 95%CI = 95% Confidence Intervals for quartile comparisons.
 JOURNAL OF SPORTS SCIENCES 5

(a) (b)

230.00
114.00
109.00 220.00

200 m Race Time (Sec)

104.00 210.00
100 m Race Time (Sec)

99.00 200.00
94.00 190.00
89.00
180.00
84.00
170.00
79.00
74.00 160.00
69.00 150.00
64.00 140.00
10 11 12 13 14 15 16 10 11 12 13 14 15 16
Chronological Age Group & Relative Age Chronological Age Group & Relative Age

Figure 1. (a & b) Curvilinear relationship between chronological (& relative) age and (a) 100 m and (b) 200 m Breaststroke swimming performance.

p = 0.03; Q1 v Q4 OR = 1.87). With the exception of 15-year-olds Part 2: Correctively adjusted distributions
in the 100 m (i.e., “Top 10%”) and 200 m events (i.e., “Top 25%”-
“10%”), RAE effect sizes increased in magnitude with selection Following adjustment of “All” individual performance times in
levels in other age-groups. In the 100 m Breaststroke, medium the 100 m and 200 m using longitudinal trendline equations,
RAE effect sizes were evident for 12–14 year-olds in the “Top the re-tabulation of relative age distributions is summarised in
25%” and “10%” of raw race times (e.g., 14 years – Top 25% Table 2. Results identified that predominantly no RAEs were
X2 = 15.61, p < 0.001, Q1 v Q4 OR = 2.38; Top 10% X2 = 11.43, apparent (i.e., Q1 distribution > Q2-Q4) across all age groups,
p = 0.01, Q1 v Q4 OR = 2.68). In the 200 m, a large RAE effect size events and selection levels (“Top 25 & 10%”) examined. The
was evident in the 13 year-olds “Top 25%” and “Top 10%” (e.g., only exception was the 13 years age groups for “Top 25%”
13 years – Top 25% X2 = 32.43, p < 0.001, Q1 v Q4 OR = 6.00; Top 100 m and 200 m performance times. There were also no
10% X2 = 20.78, p < 0.001; Q1 v Q4 OR = 44.00). Medium RAE significant OR comparisons for any particular quartile compar-
effect sizes were seen in the 12 year-olds across all selection ison in the 12, 14 and 15-year-old age groups and at any
levels. selection level. Results identified that corrective adjustments

Table 2. Relative age distribution, chi-square and odds ratio analysis of female 100 m and 200 m Breaststroke swimmers (aged 12–15 years) at state and national level
championships for 2013–2018 (inclusive) based on correctively adjusted swim times.
Performance Age- Total OR OR OR
level Group N Q1% Q2% Q3% Q4% X2 P V ES cat. Q1vQ4 (95%CI) Q2vQ4 (95%CI) Q3vQ4 (95%CI)
Top 25% 12 years 165 26.10 24.85 25.45 23.64 0.22 0.97 0.02 no 1.11 (0.60–2.04) 1.05 (0.57–1.95) 1.08 (0.58–1.99)
of 100 m
swim times
13 years 161 30.43 29.19 25.47 14.91 9.60 0.02* 0.14 small 2.04* (1.06–3.93) 1.96* (1.01–3.78) 1.71 (0.88–3.33)
14 years 159 27.67 30.19 24.53 17.61 5.65 0.13 0.11 small 1.57 (0.82–3.00) 1.71 (0.90–3.25) 1.39 (0.72–2.68)
15 years 139 30.22 28.06 20.14 21.58 4.00 0.26 0.10 small 1.40 (0.72–2.72) 1.30 (0.67–2.54) 0.93 (0.46–1.87)
Top 10% 12 years 66 24.24 24.24 33.33 18.18 3.09 0.38 0.12 small 1.33 (0.48–3.67) 1.33 (0.48–3.67) 1.83 (0.69–4.88)
of 100 m
swim times
13 years 65 26.15 26.15 23.08 24.62 0.17 0.98 0.03 no 1.06 (0.40–2.80) 1.06 (0.40–2.80) 0.94 (0.35–2.51)
14 years 64 20.31 35.94 28.13 15.63 6.12 0.10 0.18 medium 1.30 (0.44–3.81) 2.30 (0.83–6.35) 1.80 (0.64–5.08)
15 years 56 32.14 21.43 30.36 16.07 3.86 0.28 0.15 small 2.00 (0.67–5.95) 1.33 (0.43–4.16) 1.89 (0.63–5.65)
Top 25% of 12 years 111 26.13 29.73 26.13 18.02 3.27 0.35 0.10 small 1.45 (0.67–3.15) 1.65 (0.77–3.54) 1.45 (0.67–3.15)
200 m swim
times
13 years 108 34.26 30.56 25.00 10.19 14.51 0.002* 0.21 medium 3.36* (1.42–7.93) 3.00* (1.26–7.13) 2.45 (1.02–5.92)
14 years 109 23.85 27.52 30.28 18.35 3.48 0.32 0.10 small 1.30 (0.59–2.86) 1.50 (0.69–3.26) 1.65 (0.77–3.56)
15 years 97 24.74 27.83 20.62 26.80 1.18 0.75 0.06 small 0.92 (0.42–2.03) 1.04 (0.48–2.26) 0.77 (0.34–1.73)
Top 10% 12 years 45 24.44 26.67 28.89 20.00 0.78 0.85 0.08 small 1.22 (0.37–4.09) 1.33 (0.40–4.41) 1.44 (0.44–4.73)
of 200 m
swim times
13 years 44 36.36 20.45 31.82 11.36 6.72 0.08 0.23 medium 3.20 (0.87–11.81) 1.80 (0.45–7.12) 2.80 (0.75–10.47)
14 years 44 22.73 27.27 29.55 20.45 0.91 0.82 0.08 small 1.11 (0.33–3.80) 1.33 (0.40–4.44) 1.44 (0.44–4.76)
15 years 39 25.64 30.77 15.38 28.20 2.13 0.54 0.13 small 0.91 (0.26–3.12) 1.09 (0.33–3.65) 0.55 (0.14–2.07)
Q1 to Q4 = Quartile 1 to 4; Q1-Q4% = Relative age quartile (3 months combined) percentage of total number; χ2 = Chi-Square value; P = Probability value;
* = Significance p < 0.05; V = Cramer’s V effect size. ES cat. = Effect Size category; OR = Odds Ratio; 95%CI = 95% Confidence Intervals for quartile comparisons.
6 S. ABBOTT ET AL.

lead to a return of normative (expected) relative age distribu- Given that baseline analyses showed significant RAEs across all
tions (≈ 25% per quartile). To graphically summarise the age-groups and selection levels, except for 15 year-olds, find-
changes in relative age distributions from “Raw” to ings demonstrate the successful application of corrective
“Correctively adjusted” swim times, see Supplementary adjustment procedures. Only for a sub-group of 13 year-olds
Material 1 a and b which showcases data related to the (i.e., “Top 25%”) did a small-medium RAE effect size trend
12 year olds in (a) 100 m and (b) 200 m Breaststroke events. remain in the 100 m (X2 = 9.60, p = 0.02) and 200 m
(X2 = 14.50, p = 0.002) events, respectively. One likely explana-
tion for this finding relates to the initial RAE biases evident at
Discussion
baseline, particularly when considering the Q1v Q4 ORs (e.g.,
This study firstly aimed to generate accurate estimates of the OR = 6.00). Specifically, there was such a higher frequency of
relationship between decimal age and Breaststroke swimming relatively older (Q1) swimmers by proportion, and who
performance based on longitudinal competition data. achieved faster performance times, that their overrepresenta-
The second aim was to determine whether corrective adjust- tion could not be removed after corrective adjustments.
ment procedures could be utilised to remove RAEs, permitting Although not shown in present data, it should be noted that
more accurate performance evaluation. The present study was when national age-group swimmers were removed for the 13-
the first to test and apply corrective adjustment procedures in year-old 100 m sample and performance times re-examined
female youth athletes and with reference to competitive leaving only state age-group swimmers, RAEs were removed
Breaststroke swimmers. In a previous corrective adjustment after corrective adjustment (i.e., Q1 = 26.67, Q2 = 26.00,
study with male freestyle swimmers (Cobley et al., 2019), Q3 = 30.00, Q4 = 17.33, p = 0.15).
a curvilinear (quadratic) trend based on longitudinal data was The success in generating accurate longitudinal perfor-
used to summarise performance change over time. Using simi- mance estimates specific to sex, stroke and event distance
lar methodological procedures (e.g., ≥ five data-points per coupled with the effective removal of RAEs strengthens the
swimmer) although with a larger (i.e., >765) longitudinal data, validity and applicability of corrective adjustment procedures
the present study also identified significant and consistent in youth sport (Cobley et al., 2019; Romann & Cobley, 2015).
curvilinear performance trends with chronological age. Corrective adjustment procedures could have valuable implica-
Longitudinal performance profiles did vary between individuals tions for athlete training and evaluation (Till et al., 2018) and
and were accounted for in the final fixed effect estimates for evaluation in competitive (performance) contexts (Cobley et al.,
each event. Expected performance times in the 100 m 2019). For instance, corrective adjustments may benefit coach
Breaststroke event were estimated to generally reduce from education and swimmer motivation by creating awareness of
approximately 102.6 s at 10 years-old to 78.3 s at 16 years-old short-term relative-age associated performance (dis)advan-
(see Figure 1(a)) While in the 200 m Breaststroke, performance tages as well as awareness of RAE transiency in the longer-
times reduced from approximately 213.1 s at 10 years-old to term. More broadly speaking, the negative outcomes (e.g.,
166.2 s at 16 years-old (see Figure 1(b)). dropout & withdrawal) associated with early talent identifica-
The expected longitudinal performance change estimates tion and selection are also key contemporary issue for several
permitted the testing of corrective adjustment procedures, tar- sporting organisations (Cobley & Till, 2017; Delorme et al., 2009;
geting RAE removal. At the date of competition, relative age Lemez et al., 2014). Corrective adjustment procedures may
differences between 1 day to almost a year (0.99 years) were permit greater accuracy in performance evaluation within an
present in the independent swimmer sample and when swim- athlete selection context.
mers were allocated to age-groups (and afterwards relative age In Australian competitive swimming, current regional and
quartile – Q1-4). As expected, RAEs were found in the state-level talent identification systems commonly select ath-
12–14 years age groups for both events. Further, RAE effect letes based on performance times at benchmark age-group
sizes generally increased when simulated selection criteria championship events (e.g., state/national age-group rankings).
were imposed at the “Top 25%” and “Top 10%” of performance Given present results demonstrate that relative age affects
times. Relatively older females (Q1) were significantly overrepre- performance times, the use of raw performance times criteria
sented at all competition and selection levels compared to the to identify talented swimmers may be problematic. For exam-
relative younger (Q4) and when considered against the expected ple, based on registered 100 m and 200 m Breaststroke races
normal distribution. For example, nearly half the competitors in (N = 69 and 68, respectively) from state and national age-group
the 13-year-old 200 m Breaststroke with the “Top 25%” of per- competitions between 2013 and 2018 (inclusive) we re-
formance times were from Q1 (i.e., 45%). Present findings both examined and correctively adjusted performance times com-
align with prior-female swimming data, which consistently high- pared to raw performance times. The discrepancy in who made
lighted RAE prevalence in swimmers at Australian national age- the ‘Top 10ʹ rankings in each race identified that, on average,
group championships over a fifteen-year period (Cobley et al., approximately 1 in 5 swimmers (e.g., 16-20%) at 12–13 years of
2018). Likewise, findings align with studies illustrating how RAE age were different. Therefore, any strategy or procedure that
effect size increases with competition selection-level (Cobley helps address expected performance advantages and removes
et al., 2009; Smith et al., 2018; Till et al., 2010). developmental inequalities is important. That said however, the
Importantly, when corrective adjustment procedures were inclusion of additional developmental factors, such as growth
applied to all swimmers and selection groups (e.g., “Top 25%” and maturation, may more extensively account for greater
and “Top 10%”), findings identified that relative age inequalities inter-individual variability and is a recommendation for consid-
were removed, irrespective of age-group and selection criteria. eration in future work. From a research standpoint, it also
 JOURNAL OF SPORTS SCIENCES 7

remains important to determine the feasibility for how and effects in sport. Sports Medicine, 39(3), 235–256. https://doi.org/10.
when corrective adjustments could be applied in swimming 2165/00007256-200939030-00005
(e.g., training or competition) and/or other sport contexts, and Cobley, S., Till, K., O’Hara, J., Cooke, C., & Chapman, C. (2014). Variable and
changing trajectories in youth athlete development: Further verification
whether they can help attain more positive outcomes (e.g., in advocating a long-term inclusive tracking approach. Journal of
increased motivation, longer-term participation, reduce drop- Strength and Conditioning Research, 28(7), 1959–1970. https://doi.org/
out, better use of resources). 10.1519/JSC.0000000000000353
Cobley, S. P., & Till, K. (2017). Participation trends according to relative age
across youth UK Rugby League. International Journal of Sports Science &
Conclusion Coaching, 12(3), 339–343. https://doi.org/10.1177/1747954117710506
Cramer, H. (1999). Mathematical methods of statistics. Princeton University
Based on accurate longitudinal reference data summarising the Press.
relationship between decimal age (i.e., chronological and rela- Cumming, S., Searle, C., Hemsley, J., Haswell, F., Edwards, H., Scott, S.,
tive age), corrective adjustment procedures were able to Gross, A., Ryan, D., Lewis, J., White, P., Cain, A., Mitchell, S. B., &
Malina, R. M. (2019). Biological maturation, relative age and
remove RAEs from 100 to 200 m Breaststroke swimming in
self-regulation in male professional academy soccer players: A test of
a large cohort of youth female swimmers. Findings reinforce the underdog hypothesis. Psychology of Sport and Exercise, 39, 147–153.
the capability of corrective adjustment procedures to remove https://doi.org/10.1016/j.psychsport.2018.08.007
relative age-related participation and performance inequalities Deaner, R. O. (2013). Lowen A and Cobley S. Born at the wrong time:
from youth female swimming events; improve youth female Selection bias in the NHL draft. PloS One, 8(2), e57753. https://doi.org/
10.1371/journal.pone.0057753
swimming participation experiences; and, to potentially permit
Delorme, N., Boiché, J., & Raspaud, M. (2009). The relative age effect in elite
greater accuracy in swimmer evaluation. sport: The French case. Research Quarterly for Exercise and Sport, 80(2),
336–344. https://doi.org/10.1080/02701367.2009.10599568
Delorme, N., Boiché, J., & Raspaud, M. (2010). Relative age effect in female
Acknowledgments sport: A diachronic examination of soccer players. Scandinavian Journal
of Medicine & Science in Sports, 20(3), 509–515. https://doi.org/10.1111/
This research study was supported by Swimming Australia. Authors would
sms.2010.20.issue-3
like to thank their collaboration and support. Authors would also like to
Dormehl, S., Robertson, S., & Williams, C. (2016). Modelling the progression
acknowledge the assistance provided by Drew McGregor.
of male swimmers’ performances through adolescence. Sports, 4(1), 2–9.
https://doi.org/10.3390/sports4010002
Furley, P., Memmert, D., & Weigelt, M. (2016, February 2). “How much is that
Disclosure statement player in the window? The one with the early birthday?” Relative age
influences the value of the best soccer players, but not the best busi-
No potential conflict of interest was reported by the authors.
nesspeople. Frontiers in Psychology, 7, 84. https://doi.org/10.3389/fpsyg.
2016.00084
Gerdin, G., Hedberg, M., & Hageskog, C. A. (2018, June). Relative Age Effect
Funding in Swedish Male and Female Tennis Players Born in 1998–2001. Sports, 6
There was no financial assistance associated with this study. (2), 38. https://doi.org/10.3390/sports6020038
Hancock, D., Starkes, J., & Ste-Marie, D. (2015). The relative age effect in
female gymnastics: A flip-flop phenomenon. Journal of Sport Psychology,
ORCID 46(6), 714–725. doi:10.7352/IJSP.2015.46.714. http://www.ijsp-online.
com
Shaun Abbott http://orcid.org/0000-0002-6111-1033 Hancock, D. J., Adler, A. L., & Côté, J. (2013). A proposed theoretical model to
Michael Romann http://orcid.org/0000-0003-4139-2955 explain relative age effects in sport. European Journal of Sport Science, 13
Stephen Cobley http://orcid.org/0000-0001-6099-392X (6), 630–637. https://doi.org/10.1080/17461391.2013.775352
Helsen, W. F., Starkes, J. L., & Van Winckel, J. (1998). The influence of relative
age on success and dropout in male soccer players. American Journal of
References Human Biology: the Official Journal of the Human Biology Council, 10(6),
791–798. https://doi.org/10.1002/()1520-6300
Australian Bureau of Statistics. Births. Retrieved June 25, 2014, from www. Helsen, W. F., Van Winckel, J., & Williams, M. A. (2005). The relative age effect
abs.gov.au in youth soccer across Europe. Journal of Sports Sciences, 23(6), 629–636.
Brustio, P. R., Lupo, C., Ungureanu, A. N., Frati, R., Rainoldi, A., & Boccia, G. https://doi.org/10.1080/02640410400021310
(2018). The relative age effect is larger in Italian soccer top-level youth Lemez, S., Baker, J., Horton, S., Wattie, N., & Weir, P. (2014). Examining the
categories and smaller in Serie A. PloS One, 13(4), e0196253. https://doi. relationship between relative age, competition level, and dropout rates
org/10.1371/journal.pone.0196253 in male youth ice-hockey players. Scandinavian Journal of Medicine &
Cobley, S. (2016). Growth and maturation in athlete development: An educa- Science in Sports, 24(6), 935–942. https://doi.org/10.1111/sms.2014.24.
tional resource for sporting organisations. Australian Institute of Sport. issue-6
Cobley, S., Abbott, S., Dogramaci, S., Kable, A., Salter, J., Hintermann, M., & Lemez, S., Macmahon, C., & Weir, P. (2016). Relative age effects in women’s
Romann, M. (2018). Transient relative age effects across annual age groups rugby union from developmental leagues to World Cup tournaments.
in national level Australian swimming. Journal of Science and Medicine in Research Quarterly for Exercise and Sport, 87(1), 59–67. https://doi.org/10.
Sport, 21(8), 839–845. https://doi.org/10.1016/j.jsams.2017.12.008 1080/02701367.2015.1116120
Cobley, S., Abbott, S., Eisenhuth, J., Salter, J., McGregor, D., & Romann, M. López de Subijana, C., & Lorenzo, J. (2018). Relative age effect and
(2019, June). Removing relative age effects from youth swimming: The long-term success in the Spanish soccer and basketball national teams.
development and testing of corrective adjustment procedures. Journal Journal of Human Kinetics, 65(1), 197–204. https://doi.org/10.2478/hukin-
of Science and Medicine in Sport, 22(6), 735–740. https://doi.org/10.1016/ 2018-0027
j.jsams.2018.12.013 Lovell, R., Towlson, C., Parkin, G., Portas, M., Vaeyens, R., & Cobley, S. (2015).
Cobley, S., Baker, J., Wattie, N., & McKenna, J. (2009). Annual age-grouping Soccer player characteristics in English lower-league development pro-
and athlete development: A meta-analytical review of relative age grammes: The relationships between relative age, maturation,
8 S. ABBOTT ET AL.

anthropometry and physical fitness. PloS One, 10(9), 14. https://doi.org/ fitness development in adolescent athletes. Journal of Strength and
10.1371/journal.pone.0137238 Conditioning Research, 27(5), 1313–1321. https://doi.org/10.1519/JSC.
Malina, R. M., Ribeiro, B., Aroso, J., Cumming, S. P., Unnithan, V., & 0b013e31828a1ea7
Kirkendall, D. (2007). Characteristics of youth soccer players aged13-15 Till, K., Cobley, S., O’Hara, J., Cooke, C., & Chapman, C. (2014). Considering
years classified by skill level. British Journal of Sports Medicine, 41(5), maturation status and relative age in the longitudinal evaluation of
290–295. https://doi.org/10.1136/bjsm.2006.031294 junior rugby league players. Scandinavian Journal of Medicine & Science
Musch, J., & Grondin, S. (2001). Unequal competition as an impediment to in Sports, 24(3), 569–576. https://doi.org/10.1111/sms.2014.24.issue-3
personal development: A review of the relative age effect in sport. Till, K., Cobley, S., Wattie, N., O’Hara, J., Cooke, C., & Chapman, C. (2010). The
Developmental Review, 21(2), 147–167. https://doi.org/10.1006/drev. prevalence, influential factors and mechanisms of relative age effects in
2000.0516 UK Rugby League. Scandinavian Journal of Medicine & Science in Sports,
Myburgh, G. K., Cumming, S. P., Coelho, E., Silva, M., Cooke, K., & RM, M. 20(2), 320–329. https://doi.org/10.1111/j.1600-0838.2009.00884.x
(2016, October). Growth and maturity status of elite British junior tennis Till, K., Morris, R., Emmonds, S., Jones, B., & Cobley, S. (2018). Enhancing the
players. Journal of Sports Sciences, 34(20), 1957–1964. https://doi.org/10. evaluation and interpretation of fitness testing data within youth
1080/02640414.2016.1149213 athletes. Strength & Conditioning Journal, 40(5), 24–33. doi:10.1519/
Nakata, H., & Sakamoto, K. (2012). Sex differences in relative age effects SSC.0000000000000414
among Japanese athletes. Perceptual and Motor Skills, 115(1), 179–186. Vacher, P., Nicolas, M., Martinent, G., & Mourot, L. (2017). Changes of
https://doi.org/10.2466/10.05.17.PMS.115.4.179-186 swimmers’ emotional states during the preparation of national cham-
Okazaki, F., Keller, B., Fontana, F., & Gallagher, J. (2011). The relative age pionship: Do recovery-stress states matter? Frontiers in Psychology, 8,
effect among female Brazilian youth volleyball players. Research 1043. https://doi.org/10.3389/fpsyg.2017.01043
Quarterly for Exercise and Sport, 82(1), 135–139. https://doi.org/10.1080/ Vestheim, O. P., Husby, M., Aune, T. K., Bjerkeset, O., & Dalen, T. (2019).
02701367.2011.10599730 A population study of relative age effects on national tests in reading
Romann, M., & Cobley, S. (2015). Relative age effects in athletic sprinting literacy. Frontiers in Psychology, 10, 1761. https://doi.org/10.3389/fpsyg.
and corrective adjustments as a solution for their removal. PloS One, 10 2019.01761
(4), e0122988. https://doi.org/10.1371/journal.pone.0122988 Viru, A., Loko, J., Harro, M., Volver, A., Laaneots, L., & Viru, M. (1999). Critical
Smith, K. L., Weir, P. L., Till, K., Romann, M., & Cobley, S. (2018). Relative age periods in the development of performance capacity during childhood
effects across and within female sport contexts: A systematic review and and adolescence. European Journal of Physical Education, 4(1), 75–119.22.
meta-analysis. Sports Medicine, 48(6), 1451–1478. https://doi.org/10. https://doi.org/10.1080/1740898990040106
1007/s40279-018-0890-8 Wattie, N., Cobley, S., & Baker, J. (2008). Towards a unified understanding of
Till, K., Cobley, S., O’Hara, J., Chapman, C., & Cooke, C. (2013). An individua- relative age effects. Journal of Sports Sciences, 26(13), 1403–1409. https://
lized longitudinal approach to monitoring the dynamics of growth and doi.org/10.1080/02640410802233034