Skip to main content
Formal Specification

Blind Cat (Cypriot Blind Man's Buff) — Formal Specification

Cyprus · chase-tag · Confidence: low

← Back to game page

Implementation Suitability

Suitability: engine-readyFidelity: discretizedGap-fills: 5
# Straokatta — AI-optimized specification ## General idea An asymmetric pursuit and evasion game under asymmetric observation. A blindfolded Cat agent navigates a bounded 2D grid relying on noisy directional sound emissions (claps/cries) generated by evasive agents, attempting to touch and identify an evasive agent who tries to maximize proximity to the Cat without getting caught. ## Formal model - **Abstraction taken:** Bounded 2D grid M×M (e.g. 15×15). Continuous acoustic perception is discretized into acoustic emission signals that generate probabilistic distance/angle cues with additive Gaussian noise for the blindfolded Cat. - **Player/agent count:** 1 Cat agent (C), K Evasive agents (E1,…,EK), where 3≤K≤14. - **Turn model:** Simultaneous tick-based rounds. All agents submit movement and acoustic actions simultaneously; state updates and collision/tag checks resolve at tick end. ## Formal rules R1. **Initial Setup:** Cat C is placed at grid center (x0,y0)=(⌊M/2⌋,⌊M/2⌋) with orientation θ∈[0,2π) randomized. Evasive agents are initialized uniformly at random on non-center cells. R2. **Movement & Noise Generation:** Each tick, each agent selects an action from their valid action set: - Evasive agent Ei: `move(dx, dy)` where |dx|+|dy|≤1, and optionally `emit_sound(intensity)` where intensity∈{LOW,HIGH}. - Cat C: `move(dx, dy)` where |dx|+|dy|≤1, and `reach_tag(direction)`. R3. **Observation & Perception:** - Evasive agents observe the full grid state (perfect visual observability). - Cat C observes **zero** grid visual coordinates, but receives an acoustic observation vector 𝐬t={(rˆi,ϕˆi)}i∈emitting, where rˆi=ri+ϵr and ϕˆi=ϕi+ϵϕ (noisy distance and bearing to sound emitters). R4. **Boundary Constraint:** If an evasive agent steps outside grid boundary [0,M−1]2, that agent is immediately declared captured [gap-fill]. R5. **Tagging and Identification:** - If Cat executes `reach_tag(d)` into an adjacent cell containing evasive agent Ek, agent Ek is seized. - Identification check: Cat submits `identify(agent_id)`. The probability of correct identification is P(correct)=f(duration_adjacent,prior_encounters) [gap-fill]. - If correctly identified (or in direct tag mode [gap-fill]), Ek is declared caught. R6. **Round End & Role Swap:** When an evasive agent is successfully caught and identified, the round ends. Ek becomes Cat C for the subsequent round. R7. [gap-fill] If Tmax⁡=150 ticks elapse without a capture, the Cat incurs a penalty score and the round resets. ## Edge cases - **Tie at the end condition:** Not applicable — round ends on the first successful tag and identification event. - **A player with no legal move:** Agents always possess the trivial legal action `move(0, 0)` (remain stationary). - **Repetition / cycles:** Cycles (Cat wandering without contact) are bounded by Tmax⁡ in R7. - **Simultaneous claims on the same resource or position:** Evasive agents may share cells; if Cat tags a cell containing multiple evasive agents, the agent with the highest recent sound emission is tagged first [gap-fill]. - **Player elimination mid-round:** Not applicable — no elimination; round terminates upon first role transition. - **Resource exhaustion:** No consumable items; boundary space and sound channels are invariant. - **Termination guarantee:** Guaranteed by Tmax⁡ in R7 and the bounded playing arena. ## State representation - **Persistent state:** Grid dimensions M, agent roster, role assignments, cumulative scores/penalties. - **Per-tick state:** - Positions: (xc,yc) for Cat, {(xi,yi)}i=1K for evasive agents. - Active sound signals: list of active emission coordinates and intensities. - Tick counter t∈[0..Tmax⁡]. - **Observable information:** - Evasive agents: Full public information (positions of all agents). - Cat agent: Partially observable — only receives noisy acoustic signals 𝐬t and boundary collision notifications. ## Action space - For Evasive agent: `move(dx, dy)` (Δ≤1), `emit_sound(level)`, `taunt()`. - For Cat agent: `move(dx, dy)` (Δ≤1), `reach_tag(direction)`, `identify(candidate_id)`. ## Playability notes - **Single-player-vs-AI:** Excellent. Human player can play as the Blind Cat navigating an audio-only / radar interface, or as an evasive agent taunting AI Cat. - **Group games (n>4):** Reduces naturally to 1 Cat vs 1 Evasive agent with multiple stationary noise decoy emitters, or 1 Cat vs K sub-agents. - **Physical-skill games:** Physical blindfold disorientation is formalized as state uncertainty over an unobserved Markov process. - **What is lost:** Physical touch tactile sensations are replaced by discrete identity-check probabilities. ## Implementation hints Treat Cat agent as a POMDP agent maintaining a spatial belief distribution (particle filter or 2D occupancy heatmap) updated by noisy sound sensor readings. First milestone: implement 1v1 on a 10×10 grid with deterministic sound cues before introducing multi-agent acoustic clutter.

Download as JSON

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

Download JSON
Cite this page

Blind Cat (Cypriot Blind Man's Buff) (Cyprus). (n.d.). Games of Countries. https://gamesofcountries.com/games/cy-cyprus/straokatta

@misc{gamesofcountries-cy-cyprus-straokatta,
  title = {Blind Cat (Cypriot Blind Man's Buff) (Cyprus)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/cy-cyprus/straokatta}
}