Skip to main content
Formal Specification

Gravestone-Target Stone Pitching — Formal Specification

South Korea · dexterity-throw · Confidence: low

← Back to game page

Implementation Suitability

Suitability: engine-readyFidelity: probabilistic-proxyGap-fills: 1
# Biseokchigi — AI-optimized specification ## General idea A turn-based multi-stage target-demolition game where two teams alternate offensive series. Attackers must progress through an escalating 11-tier physical carrying and dropping ladder (foot → ankles → knees → thighs → belly → armpit → shoulder → chin → head → blind) to topple standing target stones. In team formats, teammates can attempt rescue strikes to revive stalled peers. ## Formal model - **Abstraction taken:** Physical gait balance, joint clamping friction, and gravity stone drops are abstracted into a 2-parameter stochastic check: Ptransit (maintaining balance while walking across the 5m zone) and Phit (accurate release onto the target). - **Player/agent count:** 2 teams (TA,TB), each with K agents (K∈[1,4]). - **Turn model:** Alternating team attack series. Within an active team series, agents act sequentially. ## Formal rules R1. **Setup:** Each agent i∈TA∪TB tracks their individual progress level Li∈{1,2,…,11} (initialized to 1). The target baseline holds K standing target stones, each assigned to an active attacker. R2. **Stage Progression Ladder:** - Stage k∈{1..11} defines difficulty constants (δtransit(k),δaim(k)): - 1 (Hand): (0.00,0.20) - 2 (Foot): (0.15,0.20) - 3 (Ankles): (0.20,0.25) - 4 (Knees): (0.15,0.20) - 5 (Thighs): (0.20,0.25) - 6 (Belly): (0.25,0.30) - 7 (Armpit): (0.10,0.20) - 8 (Shoulder): (0.25,0.25) - 9 (Chin): (0.30,0.25) - 10 (Head): (0.35,0.30) - 11 (Blind): (0.40,0.50) R3. **Action Execution:** On agent i's turn, they submit `attempt_stage(target_stone_id)`: - Locomotion check: Pdrop=δtransit(Li)/θbalance,i. If Sample1≤Pdrop, stone drops prematurely; attempt fails. - Impact check: Pstrike=(1.0−δaim(Li))⋅θaccuracy,i. If Sample2≤Pstrike, the target stone is toppled; attempt succeeds. R4. **Success & Advancement:** - On success: Target stone is toppled. Agent advances Li←Li+1. Agent may immediately take another turn if standing targets remain for their team in the current series. - On failure: Target stone remains standing. Agent i's turn ends. R5. **Team Revival (Saligi):** If an agent i failed on their stage, a remaining teammate j who successfully topples their own target may target agent i's standing stone. Toppling it revives agent i, allowing agent i to attempt their stage again in the current series. R6. **Inning Turnover:** When all agents on the attacking team have failed and no teammates remain to revive them, all target stones are reset upright, and the defending team becomes the attacking team. R7. **Win Condition:** The first team whose all K agents successfully complete Stage 11 wins the match immediately. R8. [gap-fill] Inning Limit Tie-breaker:** If neither team completes all 11 stages after Imax=10 full innings, the team with the higher sum of member levels ∑Li wins. ## Edge cases - Tie at the end condition: Handled by R8 via cumulative level sum ∑Li; sudden-death Stage 11 shootouts break identical sums. - A player with no legal move: Not applicable; `attempt_stage` is always available to active attackers. - Repetition / cycles: Bounded; agents monotonically advance through levels 1≤Li≤11 (levels never regress upon failure), and innings are capped at Imax=10. - Simultaneous claims on the same resource or position: Actions are strictly sequential within the team queue. - Player elimination mid-round; turn order repair: Teammates who finish Stage 11 act purely as revival support throwers for unfinished peers. - Resource exhaustion (pieces, tokens, space, time): Target stones are binary (standing/toppled) and reset per inning. - Termination guarantee: Monotonic progress across a finite discrete 11-level ladder with bounded innings guarantees termination. ## State representation - **Persistent state:** Team rosters, agent attributes (θbalance,θaccuracy), agent levels {Li}, current inning I. - **Per-inning ephemeral state:** Active team index, active attacker index, target stones array (Standing/Toppled), current series revival queue. - **Observable information:** Fully observable; all agent levels, stone states, and throw results are public. ## Action space - `attempt_stage(target_id)`: Parameter: `target_id` (index of standing target stone). Precondition: agent is active attacker on active team and target is standing. ## Playability notes - Single-player-vs-AI: Modeled as a 1v1 ladder race where human and AI alternate attacking single target stones. - Group games (n>4): Natively designed for 2v2 to 4v4 team play with cooperative revival mechanics. - Physical-skill games: Physical body carriage and joint stabilization are abstracted into discrete stochastic difficulty curves. - What is lost in the adaptation: The comical bodily contortions (waddling like ducks with stones clamped between knees or balancing stones on the chin) and playground cheering. ## Implementation hints Low complexity. The state space is compact: an array of integer levels Li∈[1,11] and boolean target flags. First milestone: implement the 11-stage success pipeline for 1v1, verify monotonic level progression, then implement the team revival logic.

Download as JSON

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

Download JSON
Cite this page

Gravestone-Target Stone Pitching (South Korea). (n.d.). Games of Countries. https://gamesofcountries.com/games/kr-south-korea/biseokchigi

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