Skip to main content
Formal Specification

Syrian Seven Stones — Formal Specification

Syria · dexterity-throw · Confidence: medium

← Back to game page

Implementation Suitability

Suitability: adaptation-requiredFidelity: probabilistic-proxyGap-fills: 2
# Sab'a Hajarat — AI-optimized specification ## General idea An asymmetric multi-agent field game played on a discretized 2D grid between an Attacking team (Builders) and a Defending team (Chasers). Attackers attempt to topple a central 7-stone stack and reassemble it stone-by-stone, while Defenders coordinate ball passes and throw projectiles to tag and eliminate all Attackers before the stack is completed. ## Formal model - **Abstraction taken:** Field discretized into a 15x15 coordinate grid with origin at center (0,0) where the 7-stone tower stands. Real-time motion discretized into synchronized discrete game ticks. - **Player/agent count:** 2 teams of M agents (default M=3: Team 0 = Attackers, Team 1 = Defenders). - **Turn model:** Synchronized tick-based movement; during each tick, all active agents choose an action simultaneously. ## Formal rules R1. Setup: Tower state Stower=7 (standing). Attacker agents at baseline y=−7. Defender agents positioned around the perimeter. Ball in possession of Defender 0. Team points PTS0=0,PTS1=0. R2. Phase 1 (Toppling): Active Attacker throws at tower from baseline. - Throw hit probability Ptopple=skillthrow. - If hit: Stower←0, fallen stones scattered across radius R≤2, game enters Phase 2. - If all 3 Attackers miss, teams swap roles immediately. R3. Phase 2 (Live Action Ticks): Each tick, all agents choose an action from their role set: - Attackers: `move(dx, dy)`, `place_stone()` (if adjacent to tower base (0,0) and holding a stone or retrieving from ground). - Defenders: `move(dx, dy)` (max 1 cell if holding ball), `pass_ball(defender_id)`, `throw_ball(target_coord)`. R4. Stone Rebuilding: Each successful `place_stone()` action by an active Attacker at (0,0) increments Stower←Stower+1. R5. Ball Tagging & Elimination: - A thrown ball travels in a straight line trajectory to `target_coord`. - If trajectory intersects an Attacker agent's cell at time of impact, hit probability is evaluated: Ptag=max⁡(0.2,throw_acc−dodge_skill). - On tag success, that Attacker is eliminated from the round (`active = False`). R6. Round Termination & Scoring: - If Stower==7: Team 0 scores 1 point, round ends. - If all Team 0 agents are eliminated (|Att_active|==0): Team 1 scores 1 point, roles swap. - [gap-fill] Round timer cap: If 100 ticks elapse without tower completion or full elimination, Team 1 is awarded 1 point. R7. Win condition: First team to reach 5 match points wins the game. ## Edge cases - **Tie at the end condition:** If match points tie at 4-4, a sudden-death single-throw round is executed. - **A player with no legal move:** Any immobilized or cornered agent can choose the `idle()` or `dodge_stance()` action. - **Repetition / cycles:** Bounded by the strict 100-tick round cap. - **Simultaneous claims on the same resource or position:** - If an Attacker reaches the tower at the same tick a Defender ball hits them, [gap-fill] ball resolution takes precedence (elimination evaluated before stone placement is committed). - Spatial collisions on the same grid cell are resolved by maintaining current positions (movement blocked). - **Player elimination mid-round; turn order repair:** Eliminated agents are set to inactive and omit actions for the remainder of the round. - **Resource exhaustion:** Exactly 7 stones; each placed stone remains in the tower until round reset. - **Termination guarantee:** Guaranteed by finite stone count (Stower≤7), finite roster of attackers, and the 100-tick round timeout. ## State representation - Persistent state: - `team_scores`: array `[int, int]`. - `phase`: integer ∈{1,2}. - `tower_height`: integer ∈{0..7}. - `agent_positions`: array of 2D coordinates for all 2M agents. - `agent_alive`: array of booleans. - `ball_pos`: 2D coordinate or `holder_id`. - `tick_counter`: integer ∈{0..100}. - Observable information: Fully observable 2D field state. ## Action space - Attackers: `move(direction)`, `gather_stone()`, `stack_stone()`, `dodge()`. - Defenders: `move(direction)`, `pass_ball(teammate_id)`, `throw_tag(direction)`. ## Playability notes - Single-player-vs-AI: Needs adaptation; the human controls 1 team coordinator or 1 active player with cooperative AI teammates executing flanking and passing routines. - Group games (n>4): Reduced to 3v3 discrete grid agents; multi-agent flocking and decoy behavior is well-preserved. - Physical-skill games: Real-time ball flight and evasion mapped to spatial line-of-sight intersection and probabilistic hit resolver. - What is lost in the adaptation: The physical sting of the rubber ball, frantic shouting, and spatial collision adrenaline. ## Implementation hints - Implement grid movement and ball passing as discrete state transitions. - Suggested first milestone: Model Phase 2 tick loop with 3 heuristic Attackers rushing the center while 3 heuristic Defenders pass and throw.

Download as JSON

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

Download JSON
Cite this page

Syrian Seven Stones (Syria). (n.d.). Games of Countries. https://gamesofcountries.com/games/sy-syria/sab-a-hajarat

@misc{gamesofcountries-sy-syria-sab-a-hajarat,
  title = {Syrian Seven Stones (Syria)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/sy-syria/sab-a-hajarat}
}