Skip to main content
Formal Specification

Crown and Anchor — Formal Specification

Guernsey · gambling-stake · Confidence: low

← Back to game page

Implementation Suitability

Suitability: engine-readyFidelity: directGap-fills: 1
# Couronne et Ancre (Crown and Anchor) — AI-optimized specification ## General idea A discrete multi-agent wagering game played with three 6-sided dice marked with six nautical and card symbols. Bettor agents allocate chips across the six symbol spaces; the Banker rolls the three dice, and payouts are made proportional to the number of dice matching each wagered symbol (1:1, 2:1, or 3:1), while unmatched wagers are surrendered to the Banker. The mathematical house edge is 7.87%. ## Formal model - **Abstraction taken:** None; the physical rolling of 3 dice is modeled directly as 3 independent uniform categorical samples from the set of 6 symbols: 𝒮={Crown,Anchor,Heart,Spade,Diamond,Club}. - **Player/agent count:** 1 `BANKER` agent, and N≥1 `BETTOR` agents. - **Turn model:** Two-phase synchronous round structure: 1. Wagering Phase: Bettors simultaneously submit bet vectors 𝐰i∈ℝ≥06. 2. Resolution Phase: Banker samples 3 dice outcomes and executes deterministic payouts. ## Formal rules R1. Symbol Set: 𝒮={s1,s2,s3,s4,s5,s6}={Crown,Anchor,Heart,Spade,Diamond,Club}. R2. Bankroll and Betting: - Each Bettor i possesses bankroll Bi≥0. - Each round, Bettor i declares a non-negative bet vector 𝐰i=(wi,1,…,wi,6) such that ∑j=16wi,j≤Bi. - Bettor bankroll is updated immediately: Bi←Bi−∑j=16wi,j. R3. Dice Roll: - 3 dice outcomes are sampled independently and uniformly: D1,D2,D3∼𝒰(𝒮). - The frequency count vector 𝐂=(c1,…,c6) is calculated, where cj=∑k=13𝕀(Dk==sj) and ∑j=16cj=3. R4. Payout Resolution: - For each Bettor i and each symbol j∈{1,…,6}: - If cj>0: Bettor i receives a return of (1+cj)⋅wi,j (consisting of original stake wi,j plus profit cj⋅wi,j). - If cj==0: The stake wi,j is forfeited to the Banker (net return 0). - Total return for Bettor i is Ri=∑j=16𝕀(cj>0)⋅(1+cj)⋅wi,j. - Bankroll updates: Bi←Bi+Ri. - Banker profit for the round is ΔBbank=∑i=1N(∑j=16wi,j−Ri). R5. Elimination and Termination: - If a Bettor's bankroll Bi<1 (minimum bet unit), that Bettor is eliminated (`BUST`). - [gap-fill] Match Horizon: The simulation runs for a fixed number of rounds M=50, or until all Bettors are eliminated. The winning Bettor is the agent with the highest bankroll at termination. ## Edge cases - Tie at the end condition: If multiple Bettors have equal bankroll at round M, the agent with the highest single-round profit wins; if still equal, the match is a draw. - A player with no legal move: A Bettor with Bi<1 is eliminated and submits no bets; active Bettors always have legal bets (including 𝐰i=𝟎, i.e., sitting out a round). - Repetition / cycles (and the loop-breaking rule): Bounded by the fixed round horizon M=50. - Simultaneous claims on the same resource or position: Wagers on symbols are independent; multiple Bettors can wager on the same symbol without interference. - Player elimination mid-round; turn order repair: Eliminated Bettors are removed from subsequent betting phases without affecting other players. - Resource exhaustion (pieces, tokens, space, time): Banker has infinite bankroll (or fixed pool); Bettor capital is strictly bounded. - Termination guarantee: Guaranteed by the hard round horizon M=50. ## State representation - Persistent state: - `round`: Integer ∈[1,M]. - `banker_balance`: Float / integer. - `bettor_bankrolls`: Array of floats [B1,…,BN]. - `active_status`: Array of booleans of size N. - Ephemeral state (per round): - `wagers`: Matrix N×6 of bet allocations. - `dice`: Array of 3 symbols (D1,D2,D3). - `counts`: Array of 6 integer symbol counts (c1,…,c6). - Observable information: Fully observable public state; all bankrolls, current wagers, and dice outcomes are public. ## Action space - Bettor i: - `place_bets(w)`: Submit vector 𝐰∈ℝ6 satisfying ∑jwj≤Bi and wj≥0. - Banker: - `roll_dice()`: Samples 3 categorical random variables. ## Playability notes - Single-player-vs-AI: Natural; 1 human bettor against AI banker and AI co-bettors. Excellent environment for testing bankroll management strategies (Kelly criterion, Martingale vs. fixed-fraction betting). - Group games (n>4): Easily scales to dozens of concurrent bettor agents since each bettor's resolution is mathematically decoupled. - Physical-skill games: Purely probabilistic decision environment. - What is lost in the adaptation: The auditory clatter of wooden dice in the leather cup and the lively banter of the fairground stall. ## Implementation hints - Probability distribution of matching dice cj for any single symbol j: - P(cj=0)=(5/6)3=125/216≈57.87% - P(cj=1)=3⋅(1/6)⋅(5/6)2=75/216≈34.72% - P(cj=2)=3⋅(1/6)2⋅(5/6)=15/216≈6.94% - P(cj=3)=(1/6)3=1/216≈0.46% - Expected return per unit bet on any symbol: E[R]=0⋅125216+2⋅75216+3⋅15216+4⋅1216=199216≈0.9213 (house edge = 7.87%). - First milestone: Implement vector betting and dice roll resolution with automated unit tests matching exact theoretical expectations.

Download as JSON

Download the full game corpus entry for this game as structured JSON.

Download JSON
Cite this page

Crown and Anchor (Guernsey). (n.d.). Games of Countries. https://gamesofcountries.com/games/gg-guernsey/crown-and-anchor

@misc{gamesofcountries-gg-guernsey-crown-and-anchor,
  title = {Crown and Anchor (Guernsey)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/gg-guernsey/crown-and-anchor}
}