Skip to main content
Formal Specification

Five Stones (Hungarian Knucklebones / Macskázás) — Formal Specification

Hungary · dexterity-small · Confidence: medium

← Back to game page

Implementation Suitability

Suitability: engine-readyFidelity: probabilistic-proxyGap-fills: 4
# Öt kövecs — AI-optimized specification ## General idea A turn-based progressive dexterity game where agents attempt a deterministic sequence of six increasingly complex micro-juggling figures, balancing risk vs accuracy on stone scatter patterns and scooping angles, with stochastic success probabilities derived from figure difficulty and agent skill. ## Formal model - **Abstraction taken:** The physical motor skill of tossing, ground scooping, and air catching is abstracted into a Bernoulli trial per figure attempt with success probability P(success)=f(figure_difficulty,scatter_distance,skill). - **Player/agent count:** 2 to 5 agents. - **Turn model:** Sequential rotation; an agent continues attempting sub-steps of their current figure until a failure occurs, whereupon turn passes to the next agent. ## Formal rules R1. Setup: - Sequence of figures: S=[1,2,3,4,5,6] corresponding to {`EGYES`, `KETTES`, `HARMAS`, `NEGYES`, `KAPUZAS`, `KEZHAT`}. - Each agent starts at `current_figure` =1, `current_step` =1, `score` =0. R2. Figure Definitions & Step Breakdown: - Figure 1 (`EGYES`): 4 steps (pick 1 stone four separate times). Difficulty D1=0.15. - Figure 2 (`KETTES`): 2 steps (pick 2 stones twice). Difficulty D2=0.25. - Figure 3 (`HARMAS`): 2 steps (pick 3 stones once, pick 1 stone once). Difficulty D3=0.35. - Figure 4 (`NEGYES`): 1 step (scoop all 4 stones at once). Difficulty D4=0.45. - Figure 5 (`KAPUZAS`): 4 steps (flick 1 stone through gate 4 times). Difficulty D5=0.40. - Figure 6 (`KEZHAT`): 1 scoring step (back-of-hand catch). Catch distribution: k∼Binomial(5,pcatch) where pcatch=0.6×Skill [gap-fill]. R3. Execution Trial: - For an active step with difficulty D, probability of success Psucc=min⁡(0.95,max⁡(0.05,Skilli−D)) [gap-fill]. - If step succeeds: Agent advances to next step. If all steps of the figure are completed, `current_figure` increments by 1. - In Figure 6, score increases by 10×k. Upon completing Figure 6, agent loops back to Figure 1 [gap-fill]. - If step fails: Agent commits a foul; turn ends immediately. Progress within the figure is preserved for the agent's next turn. R4. Win condition: The first agent to accumulate a score of ≥50 points (or complete 2 full cycles of all 6 figures) wins the game. ## Edge cases - Tie at the end condition: If multiple agents cross ≥50 points in the same cycle round, the agent with the higher point total wins; if identical, a sudden-death Figure 6 catch breaks the tie [gap-fill]. - A player with no legal move: Not applicable; attempting the current step is always the sole mandatory action. - Repetition / cycles: Bounded by the monotonically increasing score accumulation towards target 50. - Simultaneous claims on the same resource or position: Turns are strictly sequential; one agent interacts with the stones at any time. - Player elimination mid-round; turn order repair: No mid-round elimination; failed turns simply transfer execution token to the next player. - Resource exhaustion (pieces, tokens, space, time): Exactly 5 stones are used throughout the simulation. - Termination guarantee: Because Psucc>0 for all skill levels, every agent has positive probability of completing Figure 6 and accumulating points, guaranteeing eventual termination. ## State representation - Persistent state: `active_player: int`, `figure_progress: dict[int, int] (1..6)`, `step_progress: dict[int, int]`, `scores: dict[int, int]`. - Per-turn ephemeral state: `attempted_step: int`, `trial_success: bool`, `stones_caught: int`. - Observable information: Fully observable. All players observe current figure states, step progress, and total scores. ## Action space - `attempt_step()`: Execute the Bernoulli trial for the active figure/step. - `execute_kezat_catch()`: Execute the binomial catch trial for Figure 6. ## Playability notes - Single-player-vs-AI: Natural solitaire challenge (e.g. complete all 6 figures in fewest total turns) or turn-based race against AI bots with parameterized skill levels. - Group games (n>4): Scales seamlessly to 5 players with turn rotation. - Physical-skill games: Juggling hand-speed and finger agility are cleanly abstracted into calibrated step difficulties and skill modifiers. - What is lost in the adaptation: Tactile feeling of smooth river stones, finger skin friction, and fine hand positioning over the table. ## Implementation hints Very low complexity. Pure discrete state machine with Bernoulli trials. First milestone: Implement state machine cycling from Figure 1 through 6 with 100% success rate, then hook in the stochastic difficulty functions.

Download as JSON

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

Download JSON
Cite this page

Five Stones (Hungarian Knucklebones / Macskázás) (Hungary). (n.d.). Games of Countries. https://gamesofcountries.com/games/hu-hungary/ot-kovecs

@misc{gamesofcountries-hu-hungary-ot-kovecs,
  title = {Five Stones (Hungarian Knucklebones / Macskázás) (Hungary)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/hu-hungary/ot-kovecs}
}