Formal Specification
Knucklebone grasping — Formal Specification
Mongolia · dexterity-small · Confidence: low
← Back to game pageImplementation Suitability
Suitability: engine-readyFidelity: probabilistic-proxyGap-fills: 2
# Shagai shuurekh — AI-optimized specification
## General idea
A turn-based dexterity and risk-management game where an agent tosses a projectile token into the air, attempts to scoop up a chosen quantity of target knucklebones from a central pool with the same hand, and must catch the projectile before it hits the ground. Decision-making centers on selecting the risk-reward tradeoff of attempting greedy large-handful sweeps vs. conservative single-bone collections.
## Formal model
- **Abstraction taken:** The physical motor coordination of tossing, ground sweeping, and catching is replaced by a success probability function `P(success) = f(sweep_count, agent_dexterity)`, where attempting larger handfuls exponentially increases the failure/foul probability.
- **Player/agent count:** 2 to n players (typically 2 to 6).
- **Turn model:** Sequential discrete turns; a successful catch grants the option to sweep again or pass, while a foul terminates the turn immediately.
## Formal rules
R1. Setup: A shared pool of `N_pool` knucklebones (initial `N_pool = 40`) is placed on the playing mat. All player scores initialize to 0.
R2. On a turn, the active player chooses an action: `sweep(count)` where `1 <= count <= min(N_pool, max_hand_capacity)`.
R3. Stochastic resolution: The attempt succeeds with probability `P(success) = base_skill * exp(-k * count)`, where `k` is a difficulty scaling constant.
R4. Success outcome: If the attempt succeeds, `N_pool = N_pool - count`, the player's personal score increases by `count`, and the player may choose to attempt another `sweep()` or pass the turn.
R5. Failure/Foul outcome: If the attempt fails (projectile dropped or adjacent bone jostled), zero bones are collected, and the turn passes immediately to the next player.
R6. [gap-fill] Turn continuation cap: An agent may perform a maximum of 3 consecutive successful sweeps per turn to prevent a single dominant player from clearing the board in one turn.
R7. Pool exhaustion & elimination: When `N_pool == 0`, if multiple players remain, an equal-pot reset occurs where each non-eliminated player contributes `min_score` bones back to `N_pool`. Players with 0 bones are eliminated.
R8. Win condition: The last remaining non-eliminated player holding all knucklebones wins the game.
## Edge cases
- Tie at the end condition: Not applicable — one player eventually accumulates all knucklebones through progressive elimination rounds.
- A player with no legal move: Not applicable — as long as `N_pool > 0`, `sweep(1)` is always legal.
- Repetition / cycles: [gap-fill] Bounded by capping the number of pool reset rounds to 10; if unresolved, the player holding the strictly highest count wins.
- Simultaneous claims on the same resource: Not applicable — turns are strictly sequential.
- Player elimination mid-round: When an agent is eliminated during an equal-pot reset, they are removed from the rotation; turn order proceeds seamlessly among survivors.
- Resource exhaustion: When `N_pool == 0`, trigger R7 equal-pot reset.
- Termination guarantee: Guaranteed by monotonic transfer of knucklebones to more skilled agents and the bounded reset cap.
## State representation
- Persistent state: `N_pool` (integer, 0..40), `player_banks` (array of integers per player), `active_player_index` (0..n-1), `consecutive_sweeps_this_turn` (0..3).
- Per-turn ephemeral state: Declared `sweep_count`, binary success flag.
- Observable information: Fully observable — all player bank totals and pool size are public.
## Action space
- `sweep(count)`: Declare quantity of knucklebones to sweep in this aerial toss attempt (`1 <= count <= 5`).
- `pass_turn()`: Voluntarily end turn and preserve collected bones.
## Playability notes
- Single-player-vs-AI: Natural — 1 human vs. 1 AI agent competing in bank accumulation.
- Group games (n>4): Scalable directly up to 8 players without mechanic disruption.
- Physical-skill games: Skill parameterized via `base_skill` in `[0.7, 0.95]` and capacity decay `k`.
- What is lost in the adaptation: The tactile dexterity of finger spreading, bone collisions, and the physical thrill of catching the falling *sum*.
## Implementation hints
Extremely simple discrete state machine. Milestone 1: single-round accumulation game without elimination; Milestone 2: multi-round tournament with elimination pot resets.
Download as JSON
Download the full game corpus entry for this game as structured JSON.
Download JSONCite this page
Knucklebone grasping (Mongolia). (n.d.). Games of Countries. https://gamesofcountries.com/games/mn-mongolia/shagai-shuurekh
@misc{gamesofcountries-mn-mongolia-shagai-shuurekh,
title = {Knucklebone grasping (Mongolia)},
year = {n.d.},
publisher = {Games of Countries},
url = {https://gamesofcountries.com/games/mn-mongolia/shagai-shuurekh}
}