Skip to main content
Formal Specification

Matkot (Paddle-ball) — Formal Specification

Israel · dexterity-small · Confidence: high

← Back to game page

Implementation Suitability

Suitability: adaptation-requiredFidelity: probabilistic-proxyGap-fills: 3
# Matkot — AI-optimized specification ## General idea Matkot is a cooperative, real-time-ticked physical game where two agents aim to sustain the longest high-velocity ball rally. The strategic depth resides in balancing return accuracy and power: a hard, fast shot is harder for the partner to defend, but a high-speed rhythm is the ultimate goal. The game-theoretic task is pure co-adaptation and rhythm matching rather than zero-sum exploitation. ## Formal model - **Abstraction taken:** The physical space is discretized into a 1D line representing ball travel and player positions, and real-time flight is modeled as discrete "strokes". [gap-fill] Each stroke is defined by a pair (p,a) where p∈[0,1] is the power/velocity of the hit, and a∈[0,1] is the accuracy of the placement. - **Player/agent count:** Exactly 2 agents. - **Turn model:** Strictly alternating. Agent A plays, then Agent B, back and forth. ## Formal rules R1. **Setup:** Agent A and Agent B are placed at distance D=8 meters. Agent A starts by serving: committing a stroke (p0,a0) with low power (p0<0.3) and high accuracy (a0>0.8). R2. **Stroke Action:** On an agent's turn, they observe the incoming ball's parameters (pin,ain). The agent must select their response stroke (pout,aout). R3. **Stochastic Catch Check:** The success of catching/intercepting the incoming ball and executing the return is resolved stochastically: P(fail)=f(pin,ain)=(pin×(1−ain))×Fatigue_Factor - If the check fails, the ball is dropped, the rally ends immediately, and the final score is the count of successful consecutive strokes. R4. **Rhythm and Power Matching:** [gap-fill] To maintain a successful rally, the output power must match the partner's capabilities: - If pout>pin+0.2 (overpowering), the partner's P(fail) on the next turn is multiplied by 1.5. - If pout<0.1 (too weak), the rally is considered "dead" and terminates. R5. **Cooperative Scoring:** The score is the number of completed volleys. The agents share a single collective high score. No individual wins; they succeed or fail together. ## Edge cases - **Tie at the end condition:** Not applicable; cooperative score is a single scalar. - **A player with no legal move:** Not applicable; a player can always attempt a return stroke. - **Repetition / cycles:** The game naturally cycles during a successful rally, but physical drift, fatigue, and stochastic variance guarantee that P(fail) eventually approaches 1, ensuring the rally terminates. - **Simultaneous claims on the same resource:** Not applicable; turns are strictly alternating. - **Player elimination mid-round:** Not applicable; the rally ends immediately if either player drops the ball. - **Resource exhaustion:** Not applicable. - **Termination guarantee:** Guaranteed by R3 and R4, as the stochastic failure function guarantees eventual drop due to compounding power drift. ## State representation - **Persistent state:** - `rally_length`: Integer counting successful hits. - `highest_rally`: Record high score of the session. - **Per-turn ephemeral state:** - `incoming_power`: p∈[0,1]. - `incoming_accuracy`: a∈[0,1]. - `active_agent_id`: ID of the agent whose turn it is to strike. ## Action space - `return_stroke(power, accuracy)`: Execute a return shot. Precondition: the previous stroke was successfully resolved as "caught". ## Playability notes - **Single-player-vs-AI:** Natural and highly engaging. The human player plays against an AI partner that dynamically adjusts its play style (attacker or defender) to keep the rally alive. - **Group games (n>4):** Not applicable; Matkot is structurally a 2-agent game. - **Physical-skill games:** [gap-fill] The physical challenge is simulated by requiring agents to trade off power (which increases score but increases failure risk) against accuracy (which stabilizes the rally). - **What is lost in the adaptation:** The physical exertion, sweat, wind deflection, sand footing, and precise motor coordination of hitting a fast rubber ball on a hot beach are replaced by abstract strategic decisions of power-versus-accuracy selection. ## Implementation hints - Implement the basic state loop of alternating turns, tracking the score. - Model the stochastic resolver where high power and low accuracy compound to create a fail state. - Add an AI behavior that acts as a "Defender" (prioritizing accuracy over power) to train a reinforcement learning "Attacker" agent.

Download as JSON

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

Download JSON
Cite this page

Matkot (Paddle-ball) (Israel). (n.d.). Games of Countries. https://gamesofcountries.com/games/il-israel/matkot

@misc{gamesofcountries-il-israel-matkot,
  title = {Matkot (Paddle-ball) (Israel)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/il-israel/matkot}
}