Skip to main content
Formal Specification

Sammarinese Morra — Formal Specification

San Marino · guessing-hidden · Confidence: low

← Back to game page

Implementation Suitability

Suitability: engine-readyFidelity: directGap-fills: 2
# La morra — AI-optimized specification ## General idea A simultaneous-move, zero-sum matrix game of psychological prediction and finger arithmetic. In each rapid round, two agents simultaneously submit an integer finger extension (0..5) and an integer vocalized sum prediction (0..10). If exactly one agent's prediction matches the true sum of extended fingers, that agent scores 1 point. ## Formal model - **Abstraction taken:** Direct game-theoretic formalization. The rapid vocal and physical gesture timing is formalized as simultaneous action commitment into a private reveal buffer. - **Player/agent count:** 2 agents (Agent 1 and Agent 2). - **Turn model:** Simultaneous move rounds. ## Formal rules R1. Setup: Scores start at `score[1] = 0`, `score[2] = 0`. Target score `Win_Score = 16`. R2. On each round `r`: R2.1. Both agents simultaneously submit an action tuple `a_i = (fingers_i, call_i)`, where `fingers_i in {0, 1, 2, 3, 4, 5}` and `call_i in {0, 1, 2, ..., 10}`. R2.2. Legal action constraint: `call_i >= fingers_i` and `call_i <= fingers_i + 5` (an agent cannot call a sum mathematically impossible given their own finger choice). R2.3. If an agent submits an illegal tuple, it is treated as a void call (`call_i = null`). R3. Round resolution: R3.1. Compute true sum: `S = fingers_1 + fingers_2` (`0 <= S <= 10`). R3.2. Condition A: `call_1 == S` and `call_2 != S` -> `score[1] += 1`. R3.3. Condition B: `call_2 == S` and `call_1 != S` -> `score[2] += 1`. R3.4. Condition C: `call_1 == S` and `call_2 == S` -> Draw round (0 points awarded to both). R3.5. Condition D: `call_1 != S` and `call_2 != S` -> Miss round (0 points awarded to both). R4. Win condition: The match terminates when `max(score[1], score[2]) >= Win_Score`. R5. [gap-fill] Tie-at-target rule: If both agents are at 15 points and both achieve a scoring condition (not possible in standard 1-point rounds, but applicable in variant multipliers), play continues until one agent leads by 2 points. R6. [gap-fill] Max-round watchdog: If a match reaches 500 consecutive rounds without a winner, the agent currently leading wins; if tied, match is drawn. ## Edge cases - Tie at the end condition: Resolved via R5 and R6. - A player with no legal move: Not applicable — legal action tuples are always available every round. - Repetition / cycles: Bounded by R6 watchdog limit and positive probability of scoring on mixed-strategy Nash equilibria. - Simultaneous claims on the same resource or position: Simultaneous correct calls resolve to Condition C (draw, 0 points), per traditional rules. - Player elimination mid-round: If an agent disconnects, the active agent is awarded the remaining points to reach 16. - Resource exhaustion: Not applicable — no consumable resources. - Termination guarantee: Guaranteed by the finite Markov transition matrix with absorbing states at score 16. ## State representation - Persistent state: `score` (array of 2 ints `[score_1, score_2]`), `round_count` (int), `history` (list of previous action tuples `(a_1, a_2)`). - Per-round ephemeral state: `committed_actions` (map agent_id -> tuple). - Observable information: Actions are hidden until both commitments are locked, then fully revealed. Complete history of previous rounds is common knowledge. ## Action space - `commit_turn(fingers, call)`: Submit integer pair `fingers in [0, 5]` and `call in [fingers, fingers + 5]`. ## Playability notes - Single-player-vs-AI: Exceptional — natural testbed for sequence prediction, Markov opponent modeling, and counter-strategy learning (e.g. exploiter vs. minimax). - Group games (n>4): Supported as a King-of-the-Hill tag duel tournament. - Physical-skill games: Direct mathematical equivalence; physical gesture latency is eliminated. - What is lost: The psychological intimidation of loud, rhythmic shouting and visual feinting. ## Implementation hints Ideal matrix game. Action space per agent has size 6 * 6 = 36 valid tuples per round. Implement as a synchronous commit-and-reveal protocol.

Download as JSON

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

Download JSON
Cite this page

Sammarinese Morra (San Marino). (n.d.). Games of Countries. https://gamesofcountries.com/games/sm-san-marino/morra-sammarinese

@misc{gamesofcountries-sm-san-marino-morra-sammarinese,
  title = {Sammarinese Morra (San Marino)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/sm-san-marino/morra-sammarinese}
}