Formal Specification
The Morra (Finger-Sum Guessing Duel) — Formal Specification
France · guessing-hidden · Confidence: medium
← Back to game pageImplementation Suitability
Suitability: engine-readyFidelity: directGap-fills: 1
# Jeu de la mourre — AI-optimized specification
## General idea
A simultaneous, zero-sum game-theoretic duel where two agents simultaneously output a finger count and a guessed sum . An agent scores 1 point if and only if their guessed sum matches the exact sum of both shown fingers () and the opponent's guess does not. The game is an extreme test of non-stationary opponent modeling, mixed-strategy Nash equilibria, and psychological cycle exploitation.
## Formal model
- **Abstraction taken:** None required; the traditional game rules form a completely discrete mathematical system.
- **Player/agent count:** 2 agents in 1v1 duel.
- **Turn model:** Strictly simultaneous action selection per round.
## Formal rules
R1. Setup: Scores initialized to .
R2. Action selection: In each round , both agents simultaneously submit a joint decision tuple:
- Agent submits where and .
- Agent submits where and .
R3. Consistency constraint: An agent cannot guess a sum that is mathematically impossible given their own shown finger count ( and ).
R4. Round resolution: Compute actual sum .
- If : Agent scores 1 point ().
- If : Agent scores 1 point ().
- If : No points scored ( unchanged).
R5. Match termination: Rounds continue until . The agent reaching 16 points first wins the match immediately.
R6. [gap-fill] Turn delay: In real-time implementations, a maximum decision timeout of is enforced; failure to submit defaults to random uniform sampling over the legal action set.
## Edge cases
- Tie at the end condition: Match ends immediately when the 16th point is scored by one agent; simultaneous arrival at 16 is impossible because only one agent can score on any given round.
- A player with no legal move: Not applicable; 25 distinct legal pairs are always available every round.
- Repetition / cycles: While infinite non-scoring draw cycles are theoretically possible under identical deterministic strategies, the game terminates with probability 1 under any randomized mixed strategy because for each round.
- Simultaneous claims on the same resource or position: Handled cleanly by rule R4 (double correct guess results in a draw with 0 points awarded).
- Player elimination mid-round; turn order repair: Fixed 1v1 duel structure.
- Resource exhaustion (pieces, tokens, space, time): Infinite action space over discrete rounds; no consumable resources.
- Termination guarantee: For any non-degenerate stochastic policy where agents randomize over , the probability of a non-scoring round is . By the law of large numbers, 16 points are accumulated in finite rounds almost surely.
## State representation
- Persistent state: Match scores , history of past rounds .
- Per-round ephemeral state: Current round selections and (hidden until simultaneous reveal).
- Observable information: Current scores and full past round history are common knowledge. Current round action is strictly private until simultaneous resolution.
## Action space
- `play(fingers, guessed_sum)`:
- `fingers`: Integer .
- `guessed_sum`: Integer . Total legal action pairs per agent.
## Playability notes
- Single-player-vs-AI: Exceptionally natural, direct implementation. An AI can employ Markov chain pattern-tracking, frequency analysis, or regret-matching to detect and exploit human cognitive biases.
- Group games (n>4): Implemented as a King-of-the-Hill tag-team relay where a point loser steps down and the next teammate steps in.
- Physical-skill games: Direct strategic game; no physical motor skill abstraction required.
- What is lost in the adaptation: The auditory intimidation of shouting numbers at the top of one's lungs and the visual perception of micro-movements in the opponent's wrist tendons.
## Implementation hints
Extremely low complexity. The entire engine requires fewer than 50 lines of code. An agent action space is a matrix of choices. First milestone: implement 25-action payoff matrix and run 1v1 simulations against random and frequency-tracking opponents.
Download as JSON
Download the full game corpus entry for this game as structured JSON.
Download JSONCite this page
The Morra (Finger-Sum Guessing Duel) (France). (n.d.). Games of Countries. https://gamesofcountries.com/games/fr-france/jeu-de-la-mourre
@misc{gamesofcountries-fr-france-jeu-de-la-mourre,
title = {The Morra (Finger-Sum Guessing Duel) (France)},
year = {n.d.},
publisher = {Games of Countries},
url = {https://gamesofcountries.com/games/fr-france/jeu-de-la-mourre}
}