Skip to main content
Formal Specification

Five-Stone Jacks — Formal Specification

South Korea · dexterity-small · Confidence: low

← Back to game page

Implementation Suitability

Suitability: engine-readyFidelity: probabilistic-proxyGap-fills: 1
# Gongginori — AI-optimized specification ## General idea A turn-based multi-stage dexterity progression game where agents advance through five discrete execution levels (1-dan to 4-dan pick-ups followed by Kkeokgi flip-and-catch). The strategic dimension involves risk-managing stone scatter dispersal and choosing conservative vs. aggressive snatch trajectories during the final scoring flip. ## Formal model - **Abstraction taken:** Physical finger coordination, stone bounce kinematics, and mid-air trajectory tracking are abstracted into a 5-stage state machine. Stage completion is evaluated via success probabilities parameterized by agent dexterity θdex∈[0.5,1.5] and 2D stone scatter distances. - **Player/agent count:** 1 to 5 agents (modeled as 2 competing agents A1,A2). - **Turn model:** Strictly sequential. An active agent attempts stages sequentially until a failure occurs, retaining progress across cycles. ## Formal rules R1. **State Initialization:** Each agent tracks current stage Di∈{1,2,3,4,5} (initialized to 1) and cumulative score Si=0. Target score is Starget=50. R2. **Stage Execution:** On active turn, agent Ai attempts stage Di: - **Stages 1–4 (Pick-up Stages):** - Stage 1 (Single Pick): 4 sub-actions, Psuccess=(0.90⋅θdex)4. - Stage 2 (Pair Pick): 2 sub-actions, Psuccess=(0.85⋅θdex)2. - Stage 3 (Three-and-One): 2 sub-actions, Psuccess=(0.80⋅θdex)⋅(0.90⋅θdex). - Stage 4 (Four-Sweep): 1 sub-action, Psuccess=0.75⋅θdex. - If stage Di succeeds: Agent advances Di←Di+1 and immediately attempts the next stage on the same turn. - If stage fails: Turn ends immediately; Di remains at current level for the agent's next turn. - **Stage 5 (Kkeokgi Scoring Climax):** - Sub-step A (Back-of-Hand Catch): Number of stones landing on hand back is sampled: Kback∼Binomial(5,0.70⋅θdex). - Sub-step B (Airborne Snatch): If Kback>0, agent attempts snatch of Kback stones. Snatch success probability Psnatch=(0.80⋅θdex)Kback. - If snatch succeeds: Score awarded: Si←Si+Kback. Stage progress resets: Di←1. The agent may continue their turn starting at Stage 1. - If snatch fails (0 stones caught): Score +0, stage resets Di←1, turn ends. R3. **Turn Rotation:** Upon turn end (failure or voluntary completion of cycle), turn passes to (i(modN))+1. R4. **Win Condition:** The first agent to achieve score Si≥Starget wins the match immediately upon completing that turn. R5. [gap-fill] Target Score Overage:** If multiple agents exceed Starget in the same round cycle, the agent with the strictly higher point score wins. ## Edge cases - Tie at the end condition: Handled by R5; higher score above target wins; if tied, sudden-death 5-dan shootouts are played. - A player with no legal move: Not applicable; stage execution action is always available on turn start. - Repetition / cycles: Bounded; every successful 5-dan cycle adds positive points Kback≥1 to Si, strictly driving score monotonically toward Starget. - Simultaneous claims on the same resource or position: Not applicable; turns are strictly sequential. - Player elimination mid-round; turn order repair: No player elimination; all agents cycle sequentially. - Resource exhaustion (pieces, tokens, space, time): Fixed set of 5 stones; never consumed or removed. - Termination guarantee: Enforced by the positive drift of scores under independent Bernoulli trials with non-zero success probabilities toward finite Starget=50. ## State representation - **Persistent state:** Number of agents N, target score Starget=50, agent skill parameters {θdex,i}, cumulative scores {Si}. - **Per-turn ephemeral state:** Active agent index, current stage Di∈{1..5}, stones currently resting on back of hand Kback. - **Observable information:** Fully observable; stage levels, scores, and toss outcomes are completely public. ## Action space - `attempt_stage(stage_id)`: Parameter: `stage_id` ∈{1,2,3,4,5}. Precondition: `stage_id` equals agent's current stage Di. ## Playability notes - Single-player-vs-AI: Highly playable as a solo high-score time trial or 1v1 turn-based race against an AI. - Group games (n>4): Natively scales to N∈[2,5] players via round-robin queue rotation. - Physical-skill games: Micro-finger adjustments and hand flips are abstracted into discrete stochastic state transitions. - What is lost in the adaptation: The delicate tactile sensation of smooth plastic pellets colliding on the back of the knuckles and the rapid downward clawing snatch. ## Implementation hints Very low complexity. A clean 5-state Markov chain. First milestone: implement the 5-stage state machine and verify that average turns to reach 50 points matches empirical expectations (~10–15 turns for average skill θ=1.0).

Download as JSON

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

Download JSON
Cite this page

Gongginori (South Korea). (n.d.). Games of Countries. https://gamesofcountries.com/games/kr-south-korea/gongginori

@misc{gamesofcountries-kr-south-korea-gongginori,
  title = {Gongginori (South Korea)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/kr-south-korea/gongginori}
}