Skip to main content
Formal Specification

Warlpiri Stone-Rolling Target Game — Formal Specification

Australia · dexterity-throw · Confidence: low

← Back to game page

Implementation Suitability

Suitability: adaptation-requiredFidelity: probabilistic-proxyGap-fills: 3
# Weme — AI-optimized specification ## General idea Rolling stones at a target in a circle, with a scoring rule that creates a genuine risk decision: resting nearest scores one, but knocking the target clean out of the circle scores two — provided your own stone stays inside. A harder roll wins more and is likelier to overrun. ## Formal model - **Abstraction taken:** a roll becomes a resting position drawn around an aim point, with spread increasing in the force chosen. Collisions with the target resolve as a displacement check rather than simulated physics. - **Player/agent count:** 2–4, individually or in pairs. - **Turn model:** strictly alternating. ## Formal rules R1. A target stone sits at the centre of a circle about 2 units across. Throwing lines are 5–10 units away. R2. Players alternate turns, each rolling one stone per turn toward the target. R3. A roll is `roll(aim, force)` and resolves to a resting position drawn around the aim, with spread and overrun both increasing in force. R4. One point is scored for the stone resting closest to the target at the end of an end. R5. Two points are scored for striking the target clean out of the circle **while the rolled stone remains inside it**. Both conditions are required. R6. The game is played to 11 points. R7. [gap-fill] A stone that leaves the circle scores nothing and is removed. The source marks the thrown-versus-rolled penalty undocumented; this is the minimum rule needed to make force a real trade-off. R8. [gap-fill] The target is reset to the centre after each end. ## Edge cases - **Tie at the end condition:** two stones equidistant from the target split the point under R4 — or, if integral scoring is required, neither scores. Reaching exactly 11 simultaneously is impossible, since turns alternate. - **Player with no legal move:** cannot occur; a player on turn always has a legal roll, however poor. - **Repetition / cycles:** none. Every end awards at least one point or removes stones, and the score is monotonically increasing toward 11. - **Simultaneous claims:** turns strictly alternate, so no two stones are placed at once. Two stones resting at identical distance is handled under the tie rule above. - **Mid-round elimination and turn-order repair:** nobody is eliminated. Alternation is fixed. - **Resource exhaustion:** each player has 2–4 stones per end; exhausting them ends that player's participation in the end, not the game. - **Termination guarantee:** each end awards at least one point unless every stone leaves the circle, and R6 caps the game at 11. With a positive probability of scoring each end, the game ends with probability 1. ## State representation Persistent: circle radius, throwing line distance, stones per player, point target. Per-player: stones remaining this end, score, skill parameter. Per-turn ephemeral: resting positions of all stones this end, target position. Information set: fully observable — every stone's position is visible. ## Action space - `roll(aim_point, force)` — precondition: it is the actor's turn and stones remain this end. ## Playability notes - **Single-player-vs-AI:** natural, and a good testbed: the force choice is a clean risk/reward decision with a computable optimum given the spread model. - **Group games (n>4):** already small. Pairs reduce to individuals without changing the decision. - **Physical-skill model:** two coupled parameters — lateral spread and overrun distance, both increasing in force. Coupling them is what creates the trade-off; modelling them independently destroys it. - **What is lost:** the read of the ground. Compacted dirt and sand roll differently and a real player adjusts to the surface, which the spread model treats as fixed. ## Implementation hints Make spread and overrun both functions of force, and verify that the optimal force is interior — if the model says always roll maximum or always roll minimum, the coupling is wrong and the game is degenerate. Only two [gap-fill]s here, which makes this the most faithfully specified game in the Australian batch.

Download as JSON

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

Download JSON
Cite this page

Warlpiri Stone-Rolling Target Game (Australia). (n.d.). Games of Countries. https://gamesofcountries.com/games/au-australia/weme

@misc{gamesofcountries-au-australia-weme,
  title = {Warlpiri Stone-Rolling Target Game (Australia)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/au-australia/weme}
}