A combinatorial search experiment using tarot draw as a random oracle competing against directed search heuristics.

James Myint (Jay) — initial deposit, April 2026 Licensed under CC BY-SA 4.0

Tarot Attack treats a multi-deck tarot draw as a constraint satisfaction problem operating over an astronomically large state space. The central experimental question is:

At what complexity threshold does a tarot-draw random sampling heuristic become competitive with — or superior to — directed search?

This is not a mysticism project. The tarot deck is a theatrically motivated random oracle with known distributional properties and a culturally robust shuffling protocol. Its interest is computational, not divinatory.

A standard tarot deck contains 78 cards. The number of possible orderings of a single deck is:

78! ≈ 1.13 × 10^115

This exceeds the number of atoms in the observable universe by many orders of magnitude. Exhaustive search is not merely impractical — it is physically impossible. Any search strategy operating in this space is necessarily heuristic.

The experimental design begins with four decks (expansion planned), compounding the state space further. Ten shuffles of a 78-card deck produces effective randomization for experimental purposes given this space; the probability of returning to any prior state after ten shuffles is negligible.

Conflict detection operates exclusively on the Major Arcana (22 cards, The Fool through The World) across simultaneous draws from multiple decks. The Minor Arcana are excluded from conflict tracking.

Rationale: The Major Arcana represent named, semantically distinct positions. Conflicts — the same card appearing in draws from multiple decks simultaneously — are structurally meaningful in a way that numbered pip cards are not. This is a design choice, not a metaphysical one. It bounds the conflict detection problem to a tractable subspace (22 cards vs. 78) while preserving the interesting combinatorial properties.

Directed search heuristics (gradient descent, beam search, A*, etc.) outperform random sampling in low-to-moderate complexity search problems by exploiting local structure. However, directed search is vulnerable to local optima and scales poorly as the search landscape becomes increasingly rugged or high-dimensional.

Tarot Attack hypothesizes a crossover point: a complexity threshold beyond which the directed search advantage inverts, and random sampling — specifically, tarot-draw sampling — becomes asymptotically competitive or superior.

This intersects with known results in:

• No Free Lunch theorems (Wolpert & Macready, 1997): no search algorithm is universally superior across all problem classes
• Random restart heuristics in SAT solving: random restarts provably help in certain problem classes by escaping local optima
• Asymptotic behavior of random search in high-dimensional spaces: as dimensionality increases, the volume of space near local optima shrinks relative to the total space, reducing their trapping power

The tarot draw introduces an additional question: does the non-uniform weighting of Major Arcana conflict structure affect the crossover point relative to a purely uniform random oracle? This is the secondary experimental question.

Hand-drawn samples collapse statistically before the interesting structure emerges. The number of draws required to begin observing non-trivial conflict distribution patterns across four decks exceeds what any reasonable group of human participants could sustain. Computational simulation is the primary method. Hand drawing is preserved as a qualitative validation mechanism — useful for verifying that the simulation faithfully represents the physical draw process, not for generating primary data.

• Decks: 4 standard 78-card tarot decks (Rider-Waite or equivalent)
• Shuffles per draw: 10 (effective randomization)
• Conflict detection: Major Arcana only (22 cards)
• Comparison baseline: Directed search heuristic(s) TBD (candidate: simulated annealing over equivalent state space)
• Primary metric: Performance crossover point as a function of problem complexity
• Secondary metric: Distribution of Major Arcana conflicts — uniform vs. structured

Simulation code forthcoming in this repository.

• Conceptual framework established
• Prior art deposit (this document)
• Simulation implementation
• Baseline directed search implementation
• Initial results

Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0)

You are free to share and adapt this material for any purpose, provided you give appropriate credit and distribute any derivative works under the same license.

This document constitutes a public prior art deposit. Timestamp established by repository commit history.