Formal Specification
Mauritian Coin Pitching — Formal Specification
Mauritius · dexterity-throw · Confidence: low
← Back to game pageImplementation Suitability
Suitability: engine-readyFidelity: discretizedGap-fills: 3
# Kasié — AI-optimized specification
> **Reconstructed source.** The entry carries a provenance warning: its rules were written from
> general knowledge, not read from the two works it cites (Ministry of Arts; National Heritage Fund).
## General idea
A classic Mauritian street game (*zwe lontan*) of skill and chance played by children in village courtyards.
Each player contributes one metal coin or washer to the pool. Players pitch their coins from a 4-to-5 meter
throwing line toward a target line drawn against a stone wall. The player whose pitched coin lands closest to
the line earns first priority: they gather **all coins in the pool**, shake them, and cast them into the air
while calling "Heads" (*Face*) or "Tails" (*Pile*). All coins landing on the called face are claimed permanently.
If coins remain, the player with the second-closest pitch takes the remaining coins and repeats the call. Play
continues in pitching-proximity order until all coins are won.
## Formal model
- **Abstraction taken:** Discretized pitching distance combined with exact binomial coin-flip probability.
Continuous pitching distances are sorted into priority ranks, while subsequent coin tosses resolve with
fair binary probability.
- **Player/agent count:** 2 to 6 players in free-for-all contest (entry §Players).
- **Turn model:** Two-phase round structure: Phase 1 (Pitching for Priority) followed by Phase 2 (Sequential Coin Flipping).
## Formal rules
R1. Setup (entry §1): a target line (*la ligne*) is marked on flat ground **4 to 5 meters** from the throwing
line against a wall. Each player contributes **1 identical coin** into the round pool (pool size = N).
R2. Pitching phase (entry §2): each player pitches 1 coin toward the target line. Pitches are measured, and
players are **ranked by distance to the target line** (closest shoots first).
R3. Casting and calling (entry §3, §4): the first-ranked player gathers **all coins in the pool**, shakes them,
tosses them into the air, and calls **"Heads" (*Face*) or "Tails" (*Pile*)**.
R4. Claiming coins (entry §4): all coins that land showing the called face are **claimed and permanently
removed** by the active caster.
R5. Rotation on remaining coins (entry §5): if unclaimed coins remain in the pool, the **next closest player**
takes the remaining coins, tosses them, and calls a face. This repeats in priority order until all coins are won.
R6. Discretization of pitching accuracy:
[gap-fill: pitching distance to the line is modelled as a continuous random variable parameterized by player skill].
R7. Tie in pitching distance:
[gap-fill: players who land equidistant from the line pitch a tiebreak coin to determine priority].
R8. Win condition:
[gap-fill: the player who captures the most total coins across agreed rounds wins the overall match].
## Edge cases
- Tie at the end: unaddressed; if two players capture equal numbers of coins across a session, the entry
defines no tiebreak procedure; shared victory is standard.
- A player with no legal move: not applicable; every active player pitches in Phase 1 and flips when their
priority turn arrives.
- Repetition / cycles: not applicable; coin pool strictly decreases on every flip where >= 1 coin matches
the call; maximum flips are bounded by pool size.
- Simultaneous claims on one position: not applicable; pitching distances are strictly ordered into a
priority sequence.
- Elimination mid-round and turn-order repair: if the first caster matches all N coins on the initial toss,
the round ends immediately; subsequent players do not flip in that round.
- Resource exhaustion: the pool of active coins depletes monotonically from N to 0 during Phase 2.
- Termination guarantee: guaranteed with probability 1; because every tossed coin has a 0.5 probability of
matching the called face, the coin pool drains to zero in finite tosses almost surely.
## State representation
- Persistent state: `coins_won[p]` ∈ {0..∞}, `total_rounds_played`.
- Per-round ephemeral state: `pool_size` ∈ {0..n}, `priority_order` (list of player ids), `current_caster_idx`.
- Observable information: fully observable pitching distances, coin calls, and flip results.
## Action space
- `pitch_coin()`: player throws coin toward target line, generating a distance metric (R2).
- `call_and_toss(call_face)`: active caster selects "Heads" or "Tails" and tosses the remaining pool coins (R3).
## Playability notes
- Single-player-vs-AI: simple and fun push-your-luck and pitching-accuracy game.
- Reducing n > 4 to 2 agents: lossless; functions perfectly with 2 players.
- Physical-skill games: stone-wall pitching accuracy is combined with honest coin-toss probability.
- What is lost: the village courtyard ambiance, metal washer clinking sounds, and Mauritian Creole calls.
## Implementation hints
Model Phase 1 by drawing distance `d_p = abs(normal(0.0, sigma_p))` for each player and sorting ascending.
In Phase 2, initialize `pool = n`. For each player in sorted priority, player calls `face ∈ {0, 1}`; generate
`k ~ Binomial(pool, 0.5)` matching coins. Award `k` coins to `coins_won[p]` and update `pool -= k`. If `pool == 0`,
conclude the round. Suggested first milestone: 2-phase pitching and binomial coin-resolution simulation.
Download as JSON
Download the full game corpus entry for this game as structured JSON.
Download JSONCite this page
Mauritian Coin Pitching (Mauritius). (n.d.). Games of Countries. https://gamesofcountries.com/games/mu-mauritius/kasie
@misc{gamesofcountries-mu-mauritius-kasie,
title = {Mauritian Coin Pitching (Mauritius)},
year = {n.d.},
publisher = {Games of Countries},
url = {https://gamesofcountries.com/games/mu-mauritius/kasie}
}