Formal Specification
Mer Bean-Throwing Target Game — Formal Specification
Australia · dexterity-throw · Confidence: medium
← Back to game pageImplementation Suitability
Suitability: adaptation-requiredFidelity: probabilistic-proxyGap-fills: 1
# Kolap — AI-optimized specification
## General idea
Throw four beans at a mat 5–7 metres away; only beans landing wholly on the mat score, and a combined team total of 20 ends it. One of very few games in this catalogue with a fully attested scoring target, which makes it unusually complete.
## Formal model
- **Abstraction taken:** each throw resolves to a landing point drawn around the aim, spread set by skill and distance. The wholly-on-the-mat requirement becomes a containment test of a disc within a rectangle — the only geometry that matters.
- **Player/agent count:** two per team, seated or standing diagonally opposite; or one against one.
- **Turn model:** strictly alternating between the two teams' throwers.
## Formal rules
R1. Two mats are placed 5–7 metres apart. Two players sit or stand behind each, partners diagonally opposite.
R2. Each player has four kolaps and throws them at the opposite mat.
R3. One player throws all four; kolaps are collected; the opposing player at the other mat then throws. Play alternates thus.
R4. A throw resolves to a landing point drawn around the aim with skill-dependent spread.
R5. A kolap scores 1 only if it lands **completely** on the mat.
R6. A combined team total of 20 ends the game; that team wins.
R7. [gap-fill] A kolap landing partly on the mat scores nothing and is not re-thrown. The entry marks this undocumented; zero is the reading consistent with R5's "completely".
## Edge cases
- **Tie at the end condition:** cannot occur. Teams alternate strictly and the game ends the instant one reaches 20, so only one team can cross the threshold.
- **Player with no legal move:** a player who has thrown all four waits for collection — the normal rhythm, not a stuck state.
- **Repetition / cycles:** none. Combined score is non-decreasing and bounded by 20, so the state strictly progresses.
- **Simultaneous claims:** turns alternate and each team throws at its own mat, so two kolaps never contest a position. Overlapping landings are physically possible but do not interact under the rules.
- **Mid-round elimination and turn-order repair:** nobody is eliminated; alternation is fixed by R3.
- **Resource exhaustion:** four kolaps per player per turn, replenished at collection — a renewable resource pacing the game rather than limiting it.
- **Termination guarantee:** guaranteed. Each round scores non-negatively with positive probability, and R6 caps at 20, so the game ends with probability 1.
## State representation
Persistent: mat dimensions, separation (5–7 m), kolap size, target score 20.
Per-team: combined score.
Per-player: kolaps remaining this turn, skill parameter.
Per-turn ephemeral: landing points of the current throws.
Information set: fully observable.
## Action space
- `throw(aim_point)` — precondition: it is the actor's turn and kolaps remain this turn.
## Playability notes
- **Single-player-vs-AI:** natural, and cleanly optimal — aim at the mat centre; the only variable is skill. A good calibration testbed precisely because the optimal policy is known in advance.
- **Group games (n>2):** the 2v2 partnership adds no interaction — partners pool scores — so it reduces to 1v1 losing nothing structural.
- **Physical-skill model:** a bivariate spread whose variance falls with skill. R5's containment test means spread matters far more than under a nearest-to-target rule.
- **What is lost:** little of the mechanics — this is among the most faithfully specifiable games here, with one gap-fill. What is lost is the setting: partners sitting diagonally across two mats is sociable, and the spec renders it as two independent throwing lanes.
## Implementation hints
R5's all-or-nothing containment makes this behave quite unlike a nearest-to-target game: expected score is highly sensitive to spread, so a small skill gap produces a large score gap. Verify that before tuning. With one gap-fill and an attested target score, this is the best first implementation in the Australian set.
Download as JSON
Download the full game corpus entry for this game as structured JSON.
Download JSONCite this page
Mer Bean-Throwing Target Game (Australia). (n.d.). Games of Countries. https://gamesofcountries.com/games/au-australia/kolap
@misc{gamesofcountries-au-australia-kolap,
title = {Mer Bean-Throwing Target Game (Australia)},
year = {n.d.},
publisher = {Games of Countries},
url = {https://gamesofcountries.com/games/au-australia/kolap}
}