Formal Specification
Albanian Traditional Belt Wrestling — Formal Specification
Albania · wrestling-combat · Confidence: low
← Back to game pageImplementation Suitability
Suitability: adaptation-requiredFidelity: probabilistic-proxyGap-fills: 3
# Mundja me breza — AI-optimized specification
## General idea
A two-agent combat duel where both agents maintain a persistent bilateral belt grip and simultaneously select from a set of traditional throwing techniques, stance counters, and balance maneuvers each discrete round. Technique interactions resolve through a skill-weighted probability matrix to trigger clean falls, balance loss, or foul accumulations.
## Formal model
- **Abstraction taken:** Continuous physical grappling and dynamic friction on grass are discretized into rounds of simultaneous technique selection, resolved via a stochastic matchup table.
- **Player/agent count:** 2 agents (`WRESTLER_A`, `WRESTLER_B`).
- **Turn model:** Simultaneous discrete action selection per round, followed by centralized stochastic resolution.
## Formal rules
R1. Setup: Both agents start standing with `balance = 100`, `fouls = 0`, `grip_intact = true`.
R2. Round Structure: Each round, both agents simultaneously select one technique from the legal action set:
- `HIP_THROW` (High power, vulnerable to counter-trips)
- `LEG_TRIP` (Fast, vulnerable to lifts)
- `SHOULDER_LIFT` (Counters leg trips, vulnerable to hip throws)
- `DEFENSIVE_STANCE` (Reduces opponent fall probability by 50%, restores `+15` balance)
R3. Technique Resolution Matrix [gap-fill: rock-paper-scissors matchup dynamics]:
- `HIP_THROW` beats `SHOULDER_LIFT` (+30% throw success)
- `SHOULDER_LIFT` beats `LEG_TRIP` (+30% throw success)
- `LEG_TRIP` beats `HIP_THROW` (+30% throw success)
R4. Fall Calculation: Base throw success probability `P_throw = max(0.05, 0.35 + 0.003 * (balance_attacker - balance_defender) + matchup_bonus)` [gap-fill].
- If throw succeeds: Opponent suffers a clean fall (`BACK_ON_GRASS`). The thrower wins the bout immediately.
- If throw fails: Defender loses `15` balance points; attacker loses `10` balance points.
R5. Fouls and Grip Release:
- If an agent chooses `DEFENSIVE_STANCE` 3 times consecutively, they incur a foul for passive stalling (`foul += 1`) [gap-fill].
- If an agent reaches `fouls == 3`, they are disqualified and the opponent wins immediately.
R6. Draw Resolution: If after `R_max = 20` rounds no clean fall occurs, the agent with strictly higher remaining balance wins; if balance is equal, the agent with fewer fouls wins.
## Edge cases
- Tie at the end condition: Resolved by R6 balance comparison and foul count; if perfectly tied, a sudden-death single round is run.
- A player with no legal move: Not applicable; all 4 techniques are legal every round.
- Repetition / cycles: Bounded by R6's 20-round hard cutoff and balance attrition.
- Simultaneous claims on the same resource or position: Actions are committed simultaneously and evaluated through the symmetric matrix in R3.
- Player elimination mid-round; turn order repair: Not applicable; 1v1 bout terminates immediately upon fall or disqualification.
- Resource exhaustion (pieces, tokens, space, time): Not applicable; balance is bounded `[0..100]`.
- Termination guarantee: why the game provably ends: Guaranteed because `P_throw > 0` every round, balance monotonically decreases on unresolved attacks, and R6 enforces a maximum 20-round limit.
## State representation
- Persistent state: `balance: array[2] of float`, `fouls: array[2] of int`, `consecutive_defenses: array[2] of int`, `round_num: int`.
- Per-turn ephemeral state: `selected_technique: array[2] of TechniqueEnum (hidden until reveal)`.
- Observable information: Balance, fouls, and round count are public. Current-round technique choice is private until simultaneous reveal.
## Action space
- `hip_throw()`: Execute hip throw from belt grip.
- `leg_trip()`: Execute trip technique.
- `shoulder_lift()`: Lift opponent from sash hold.
- `defensive_stance()`: Maintain balance and recover stamina.
## Playability notes
- Single-player-vs-AI: Natural; human/agent plays 1v1 against an AI bot employing mixed Nash equilibrium strategies.
- Group games (n>4): Modeled as a single-elimination tournament bracket of independent 1v1 bouts.
- Physical-skill games: Physical power, grip torque, and center of gravity are abstracted into balance metrics and technique probabilities.
- What is lost in the adaptation: The tactile feel of woolen sash tension, instantaneous balance adjustments, and foot friction on dewy meadow grass.
## Implementation hints
Low state complexity (balance float + foul count per agent). Branching factor is 4 per round. First milestone: Implement simultaneous action reveal, matchup matrix lookup, and throw probability resolution for a single 1v1 match.
Download as JSON
Download the full game corpus entry for this game as structured JSON.
Download JSONCite this page
Albanian Traditional Belt Wrestling (Albania). (n.d.). Games of Countries. https://gamesofcountries.com/games/al-albania/mundje-me-breza
@misc{gamesofcountries-al-albania-mundje-me-breza,
title = {Albanian Traditional Belt Wrestling (Albania)},
year = {n.d.},
publisher = {Games of Countries},
url = {https://gamesofcountries.com/games/al-albania/mundje-me-breza}
}