Skip to main content
Formal Specification

Goat Dragging — Formal Specification

Afghanistan · animal-contest · Confidence: low

← Back to game page

Implementation Suitability

Suitability: adaptation-requiredFidelity: probabilistic-proxyGap-fills: 5
# Buzkashi — AI-optimized specification ## General idea A high-intensity, team-based equestrian possession and race game. Mounted agents compete across a zoned field to hoist a heavy carcass from the ground, defend possession against fierce physical grappling from opponent riders, round a distant turning marker (*bayraq*), and gallop back to deposit the carcass cleanly inside the scoring goal circle (*Dayira-ye Halal*). ## Formal model - **Abstraction taken:** The continuous high-speed movement and physical wrestling on horseback is abstracted into a 2D discrete zone grid (X∈[0,8],Y∈[0,4]) representing the 400m x 200m field. Continuous momentum, whip-handling, and grip mechanics are replaced by discrete movement points, horse stamina pools, and stochastic grapple/hoist resolution. - **Player/agent count:** 2 teams (Team A and Team B), with 4 to 6 active agents per team on the field. - **Turn model:** Simultaneous-tick execution (each tick represents approximately 3–5 seconds of real-time play; regulation match is 60 ticks per half, 2 halves total = 120 ticks). [gap-fill: match length formalized as 120 discrete ticks]. ## Formal rules R1. **Setup:** The field is an 8×4 grid. - Center circle (*Dayira-ye Halal*): Zone `(4, 2)`. - Starting position of carcass: Ground in Zone `(4, 2)`. - Team A turning flag (*bayraq*): Zone `(0, 2)`. - Team B turning flag (*bayraq*): Zone `(8, 2)`. - Team A agents start in Zones `(2, y)`, Team B agents in Zones `(6, y)`. R2. **Agent Attributes:** Each mounted agent possesses `speed` (1..3 movement units/tick), `strength` (1..10 grappling rating), and `stamina` (0..100). [gap-fill: numerical attribute baselines]. R3. **Turn Actions:** Every tick, all agents simultaneously choose one action: - `MOVE(dx, dy)`: Travel to adjacent zone up to current `speed`. - `HOIST()`: Attempt to lift carcass from ground (valid only if co-located with unpossessed carcass). Success probability P(hoist)=strength10×stamina100. - `GRAPPLE(target_agent)`: Attempt to wrench carcass from current possessor in same zone. - `RACE(direction)`: Sprint with carcass (+1 speed, -10 stamina). - `PASS(teammate_agent)`: [gap-fill: In modern codified team play, toss carcass to an adjacent teammate]. - `DROP_INTO_GOAL()`: Release carcass into `Dayira-ye Halal` goal. R4. **Grapple Resolution:** If one or more opponents grapple the carcass carrier in the same zone, the engine resolves a contest: ΔS=Carrier_Strength+Teammate_Support−(∑Opponent_Strength)+𝒩(0,2) - If ΔS>0, the carrier retains possession. - If ΔS<−3, possession transfers to the strongest grappler. - If −3≤ΔS≤0, the carcass drops to the ground in the current zone (*melee fumble*). R5. **Scoring Circuit Requirement:** To score a valid point, the carcass must: 1. Be possessed and carried by Team A around Flag B `(8, 2)` (or Team B around Flag A `(0, 2)`). 2. Carried back to the home scoring circle `(4, 2)`. 3. Released via `DROP_INTO_GOAL()` while inside `(4, 2)` without being fumbled. R6. **Fouls and Penalties:** Whipping opponent horses or unseating riders illegally triggers a foul penalty: offending agent is benched for 10 ticks, creating a numerical power play. [gap-fill: foul penalty duration formalized]. R7. **Match Termination & Win Condition:** The match terminates after exactly 120 ticks. The team with the higher score wins. ## Edge cases - Tie at the end condition: If scores are tied at tick 120, a sudden-death overtime period (20 ticks) begins from center setup; the first team to score wins. If still tied after overtime, the team with fewer committed fouls wins. [gap-fill: 20-tick sudden death overtime rule]. - A player with no legal move: Not applicable; an agent always has a valid `MOVE(0, 0)` (hold position) or tactical repositioning action. - Repetition / cycles: The game is strictly bounded by the 120-tick clock; cyclical carcass possession changes cannot prolong the game indefinitely. - Simultaneous claims on the same resource or position: When multiple agents attempt `HOIST()` on the carcass simultaneously on the same tick, the contest resolver (R4) treats all candidates as simultaneous contenders and awards the hoist to the highest roll. - Player elimination mid-round; turn order repair: If an agent is fouled out or injured, the team plays down an agent until the penalty expires or an authorized substitution occurs between halves. Simultaneous tick execution requires no turn order repair. - Resource exhaustion (pieces, tokens, space, time): Match clock (120 ticks) strictly exhausts. Stamina recovers partially during rest ticks but does not block legal actions. - Termination guarantee: Provably terminates because the simulation clock decrements monotonically each tick to the terminal state `tick == 120`. ## State representation - **Persistent state:** - `tick_counter`: integer (1..120) - `scores`: `{team_a: integer, team_b: integer}` - `carcass_state`: `{holder: agent_id | None, zone: (x, y), rounded_flag: boolean}` - `agent_states`: list of `{id, team, zone: (x,y), stamina: 0..100, benched_ticks: integer}` - **Per-turn ephemeral state:** - `agent_actions`: map of `agent_id -> Action` - `grapple_outcomes`: list of resolved contest deltas - **Observable information:** Fully observable to team coordinators, or localized fog-of-war where agents observe exact opponent stats only within their zone and adjacent zones. ## Action space - `move(target_x, target_y)`: Precondition: |x−target_x|+|y−target_y|≤speed. - `hoist()`: Precondition: Carcass unheld and `carcass.zone == agent.zone`. - `grapple(target)`: Precondition: `carcass.holder == target` and `target.zone == agent.zone`. - `drop_into_goal()`: Precondition: `carcass.holder == agent` and `agent.zone == (4, 2)` and `carcass.rounded_flag == true`. ## Playability notes - Single-player-vs-AI: Excellent as a team-tactics simulator where the player commands the primary carrier or issues macro-tactical orders (e.g., "Screen", "Sprint", "Support Melee") to AI teammates. - Group games (n>4): Standard match model uses 4v4 or 6v6. Reduces to a 2-agent duel (1v1 mounted race/grapple) on a condensed 5×3 grid without altering core circuit and possession rules. - Physical-skill games: Horsemanship and grappling leverage are modeled via agent `strength` and dynamic `stamina` expenditure. - What is lost in the adaptation: The visceral danger, horse psychology, footing terrain hazards, and raw physical pain of whipping and galloping collisions are transformed into clean probability distributions. ## Implementation hints Medium complexity due to multi-agent simultaneous action resolution. The grid-based zone model avoids continuous vector collision math while retaining tactical depth (flanking, blocking, lead-out galloping). Suggested first milestone: implement carcass movement, flag check, and goal scoring for 1v1 without grappling, then add team agents and the contest resolver.

Download as JSON

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

Download JSON
Cite this page

Goat Dragging (Afghanistan). (n.d.). Games of Countries. https://gamesofcountries.com/games/af-afghanistan/buzkashi

@misc{gamesofcountries-af-afghanistan-buzkashi,
  title = {Goat Dragging (Afghanistan)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/af-afghanistan/buzkashi}
}