Skip to main content
Formal Specification

Five Knucklebones — Formal Specification

Kazakhstan · dexterity-small · Confidence: low

← Back to game page

Implementation Suitability

Suitability: adaptation-requiredFidelity: probabilistic-proxyGap-fills: 4
# Bes asyq — AI-optimized specification ## General idea Bes asyq is a sequential tabletop jacks game played with five sheep astragalus bones. Players take turns attempting a rigorous 7-stage sequence of escalating manual dexterity figures—tossing a master bone (*khan*), scooping or flicking ground bones in prescribed quantities and geometric patterns, and catching the falling *khan* with a single hand. A single fault ends the player's turn immediately, with progress preserved for subsequent turns until one player completes the entire sequence. ## Formal model - **Abstraction taken:** Manual dexterity (aerial toss height, hand inversion, fingertip flick precision, and hand-eye reaction timing) is abstracted into a parameterized stochastic success function. [gap-fill] For each stage s∈{1..7} and sub-action, success probability is modeled as P(success)=σ(βs⋅(skill−difficultys)), where skill∈[0,1] is the agent's dexterity parameter and difficultys increases monotonically with stage complexity. - **Player/agent count:** 2 to 4 agents in sequential rotation (standard 2-agent duel default). - **Turn model:** Sequential single-player active turns; active agent executes sub-actions until a fault occurs or the game is won. ## Formal rules R1. **Game Setup and State Initialization:** - Pieces: 5 knucklebones (1 designated master *khan*, 4 standard knucklebones). - Each agent i∈{1..N} has a persistent progress state: `stage` ∈{1..7} (initially 1) and `sub_step` ∈{0..Ks} (initially 0). - Turn order is fixed in round-robin order: 1→2→⋯→N→1. R2. **The 7-Stage Progression Sequence:** - **Stage 1 (Birlik / Singles):** 4 sub-steps. Scatter bones; for each sub-step k∈{1..4}, agent tosses *khan*, scoops 1 ground bone, and catches *khan*. - **Stage 2 (Ekilik / Doubles):** 2 sub-steps. Scatter bones; for each sub-step k∈{1..2}, agent tosses *khan*, scoops 2 ground bones in one grab, and catches *khan*. - **Stage 3 (Üshtik / Triples):** 2 sub-steps. Scatter bones; sub-step 1 scoops 3 ground bones together; sub-step 2 scoops the remaining 1 ground bone. - **Stage 4 (Törttik / Quadruples):** 1 sub-step. Cluster 4 bones together; toss *khan*, scoop all 4 ground bones simultaneously, and catch *khan*. - **Stage 5 (Qarshu / Clutching):** 2 sub-steps. Hold all 5 bones in palm, toss all 5 up, catch ≥1 on back of hand (sub-step 1); toss from back of hand and catch all in palm (sub-step 2). [gap-fill] Dropping any bone during either catch constitutes a fault. - **Stage 6 (Köpir / The Bridge):** 4 sub-steps. Non-active hand forms an arch on table; scatter 4 bones in front; for each sub-step k∈{1..4}, toss *khan*, flick 1 bone through the arch without touching the arch, and catch *khan*. - **Stage 7 (Qaqpaqyl / Final Juggles):** 1 sub-step. Rapid aerial juggling catch sequence of all 5 bones. R3. **Turn Execution & Fault Handling:** - On an agent's turn, the agent attempts the next sub-action required by their current `(stage, sub_step)`. - If the stochastic check succeeds: - `sub_step` increments. - If `sub_step` reaches the stage ceiling Ks, the agent advances to `stage` ←stage+1, `sub_step` ←0. - The agent continues playing their turn immediately. - If the stochastic check fails (a fault / *janu* occurs): - The agent's turn terminates immediately. - Current `(stage, sub_step)` is saved without regression. - Turn passes to the next agent in cyclic rotation. R4. **Winning Condition:** - The first agent to successfully complete Stage 7 (`stage == 7` and `sub_step == 1`) immediately wins the game, and the match terminates. ## Edge cases - **Tie at the end condition:** Not applicable; turns are strictly sequential, and the first agent to clear Stage 7 triggers immediate victory. - **A player with no legal move:** Not applicable; an active agent always has a mandatory skill action attempt corresponding to their current stage. - **Repetition / cycles (and the loop-breaking rule):** Since player progress never regresses upon a fault (stages are monotonically non-decreasing) and P(success)>0 for all stages, infinite loops without progression have probability 0. [gap-fill] A global turn cap of 200 total turns ends the match in favor of the player with the highest `(stage, sub_step)` progress. - **Simultaneous claims on the same resource or position:** Not applicable; single active player at a time. - **Player elimination mid-round; turn order repair:** If an agent resigns or times out, they are removed from the active players list, and the turn pointer advances to `(current_index + 1) mod remaining_count`. - **Resource exhaustion (pieces, tokens, space, time):** The 5 bones are an invariant physical set, reset on each stage attempt. - **Termination guarantee: why the game provably ends:** Player state `(stage, sub_step)` is strictly non-decreasing and bounded above by Stage 7. Combined with the 200-turn safety ceiling [gap-fill], termination is mathematically guaranteed. ## State representation - **Persistent state:** - `num_players`: Integer N∈[2,4]. - `active_player_idx`: Integer 0..N−1. - `player_progress`: Array of N structs `[{stage: 1..7, sub_step: 0..4}, ...]`. - `turn_count`: Total turns taken. - **Per-turn ephemeral state:** - `current_attempt_stage`: Stage being executed. - `attempt_success`: Boolean outcome of the stochastic resolver. - **Observable information:** Perfect information. All player progress stages and attempt outcomes are fully public. ## Action space - `attempt_figure(stage, sub_step)`: - Parameters: Implicitly matches the agent's current saved `(stage, sub_step)`. - Preconditions: It is the agent's active turn. ## Playability notes - **Single-player-vs-AI:** Natural. One human player vs 1–3 AI bots, each rolling against their respective skill parameters. - **Group games (n>4):** Reduces directly to 2 agents without losing any core mechanic, as player interaction is purely turn-rotation race competition. - **Physical-skill games:** Skill parameter S∈[0,1] modulates the sigmoid success probability across the 7 stages. - **What is lost in the adaptation:** Tactile bone balancing on the back of the hand, geometric spatial scattering on the mat, and physical finger-flicking tension under the hand bridge. ## Implementation hints - Model stages as a state machine where each node has a success transition to next sub-step and a failure transition that yields turn execution to next player. - Set stage difficulty parameters such that early stages (Birlik, Ekilik) have low failure rates (∼0.1), while late stages (Qarshu, Köpir, Qaqpaqyl) have higher failure rates (∼0.4−0.6). - First milestone: Implement the 7-stage state machine with deterministic success, verify stage transitions, then attach the stochastic failure resolver.

Download as JSON

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

Download JSON
Cite this page

Five Knucklebones (Kazakhstan). (n.d.). Games of Countries. https://gamesofcountries.com/games/kz-kazakhstan/bes-asyq

@misc{gamesofcountries-kz-kazakhstan-bes-asyq,
  title = {Five Knucklebones (Kazakhstan)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/kz-kazakhstan/bes-asyq}
}