Formal Specification
Tranta (Romanian Traditional Belt Wrestling) — Formal Specification
Romania · wrestling-combat · Confidence: low
← Back to game pageImplementation Suitability
Suitability: engine-readyFidelity: probabilistic-proxyGap-fills: 2
# Trânta — AI-optimized specification
## General idea
A discrete, two-agent grappling contest where combatants maintain a mandatory mutual belt grip. In each exchange, agents secretly select offensive throwing techniques, defensive stances, or counter-trips from an enumerated technique set. A stochastic resolver parameterized by agent balance, fatigue, and matchup interactions computes takedowns, points, and back-fall victories (*tuș*).
## Formal model
- **Abstraction taken:** Continuous physical grappling and isometric force exchanges are discretized into sequential combat rounds (exchanges). Each exchange resolves simultaneous technique selections via a probabilistic outcome matrix.
- **Player/agent count:** Exactly 2 agents (Agent 0 and Agent 1).
- **Turn model:** Simultaneous action selection per round (max 20 rounds / 4 minutes simulation).
## Formal rules
R1. State tracking: Per agent : `points` , `stamina` , `balance` , `fouls` .
R2. Technique Set :
- `hip_throw`: High power offensive throw (requires balance , consumes 15 stamina).
- `chest_suplex`: High risk/reward throw (consumes 25 stamina).
- `leg_trip`: Fast balance disruption (consumes 10 stamina).
- `brace_defense`: Defensive balance recovery (restores balance, consumes 5 stamina).
- `counter_throw`: Reactive counter aiming to reverse aggressive enemy throws.
R3. Exchange Resolution:
- Both agents simultaneously choose .
- [gap-fill] Takedown success probability is derived from the technique interaction matrix modified by balance differential and stamina ratio.
- Outcome classes:
1. **Clean Fall (*Tuș*):** Throw lands flat on back. Bout ends immediately in victory for attacker.
2. **High-Amplitude Throw:** Attacker scores 4 points; defender balance drops by 30.
3. **Takedown:** Attacker scores 2 points; defender balance drops by 15.
4. **Standoff / Neutral:** 0 points; both agents incur stamina decay.
R4. Out-of-Bounds & Fouls:
- If an agent's balance drops to 0, they step out of bounds (opponent awarded 1 point; balance resets to 50).
- Releasing the mandatory grip without a valid attack increments `fouls += 1`.
- If `fouls == 3`: Agent is disqualified; opponent wins immediately.
R5. Win Condition:
- Immediate Win via *Tuș* (clean fall) or Disqualification (3 fouls).
- Superiority Win: A lead of points at the end of any round.
- Time Limit Win: Agent with higher point score at round 20 wins.
R6. [gap-fill] Tiebreaker: If tied at round 20, a 1-round golden-score sudden death round with zero defensive stance permitted resolves the bout.
## Edge cases
- Tie at the end condition: Resolved by R6 sudden-death golden-score round.
- A player with no legal move: `brace_defense` is always a valid default action.
- Repetition / cycles: Bounded — rounds strictly increment from 1 to 20.
- Simultaneous claims on the same resource or position: Simultaneous action selection is resolved via the joint symmetric payoff matrix.
- Player elimination mid-round; turn order repair: Not applicable (strictly 2 agents).
- Resource exhaustion: Stamina is bounded in ; stamina exhaustion forces defensive actions only.
- Termination guarantee: Enforced by the finite 20-round time horizon and absorbing *tuș* states.
## State representation
- Persistent state:
- `points`: Array of 2 integers `[p0, p1]`.
- `stamina`: Array of 2 floats `[s0, s1]`.
- `balance`: Array of 2 floats `[b0, b1]`.
- `fouls`: Array of 2 integers `[f0, f1]`.
- `round`: Integer .
- Per-round ephemeral state:
- `actions`: Tuple chosen secretly by agents.
- Observable information: Stamina, balance, points, and fouls are public knowledge. Chosen techniques are private until simultaneous reveal.
## Action space
- `select_technique(tech_id)`: `tech_id` .
## Playability notes
- Single-player-vs-AI: Highly compelling 1v1 combat simulator (rock-paper-scissors dynamic coupled with stamina and balance management).
- Group games (n>4): Not applicable (1v1 bouts; tournament bracket format).
- Physical-skill games: Physical body-weight mechanics and leverage converted into stamina, balance scalars, and technique matchups.
- What is lost: Sensory grappling touch, sweat, and the physical roar of the village fair.
## Implementation hints
- Implement as a discrete turn-based Markov Decision Process.
- Balance matrix calibration: ensure `counter_throw` beats `chest_suplex`, `chest_suplex` beats `brace_defense`, and `brace_defense` counters `hip_throw`.
Download as JSON
Download the full game corpus entry for this game as structured JSON.
Download JSONCite this page
Tranta (Romanian Traditional Belt Wrestling) (Romania). (n.d.). Games of Countries. https://gamesofcountries.com/games/ro-romania/tranta
@misc{gamesofcountries-ro-romania-tranta,
title = {Tranta (Romanian Traditional Belt Wrestling) (Romania)},
year = {n.d.},
publisher = {Games of Countries},
url = {https://gamesofcountries.com/games/ro-romania/tranta}
}