Skip to main content
Formal Specification

Blind Old Woman (Blind Man's Buff) — Formal Specification

Czechia · chase-tag · Confidence: high

← Back to game page

Implementation Suitability

Suitability: engine-readyFidelity: discretizedGap-fills: 3
# Slepá bába — AI-optimized specification ## General idea An asymmetrical pursuit-evasion game played on a bounded 2D grid where one agent (the seeker) suffers complete visual sensory occlusion, receiving only noisy directional acoustic/tactile cues generated by evaders moving within the perimeter. Evaders maneuver to avoid capture within a closed arena while generating periodic sound pulses, and the seeker must infer the evaders' positions, execute intercept trajectories, and correctly identify or tag an opponent. ## Formal model - **Abstraction taken:** Continuous physical room and blindfold navigation are mapped to an N×N discrete grid arena (default 10×10). Continuous auditory teasing is discretized into per-tick acoustic emission events that propagate with Manhattan distance decay and noise. - **Player/agent count:** 1 Seeker agent (S), n Evader agents (E1,…,En, default n=4). - **Turn model:** Simultaneous tick-based execution. Each tick, all agents choose an action simultaneously; movements and sensory emissions resolve in a fixed order. ## Formal rules R1. Setup: Arena is a bounded grid G={(x,y)|1≤x,y≤10}. Seeker S starts at center (5,5) with orientation randomized. Evaders Ei start at random distinct boundary/interior coordinates (xi,yi)≠(5,5). R2. In each tick t, every agent submits an action: - Evader: `move(dx, dy)` with |dx|+|dy|≤1, and optional `tease(loudness)` ∈{0,1,2}. - Seeker: `step(dx, dy)` with |dx|+|dy|≤1, and optional `reach(direction)` ∈{N, S, E, W, HERE}. R3. Movement resolution: Evaders move first. If an evader moves outside G, they immediately incur a penalty and become the new Seeker for the next round. Seeker moves second. R4. Sensory observation update: - Seeker receives no direct vision of coordinates. Seeker receives an array of acoustic vectors a→i=Dir(Ei→S)+ϵ where noise ϵ∼𝒩(0,σ2) scales inversely with Ei's tease loudness and directly with distance. - Evaders receive full coordinate vision of S and all other evaders. R5. [gap-fill] Mandatory Acoustic Generation: Every evader must emit a tease signal with loudness ≥1 at least once every 3 ticks to prevent silent camping. R6. Capture condition: If S occupies the same cell as an evader Ek, or executes `reach(dir)` targeting the exact cell of Ek, Ek is captured. R7. [gap-fill] Identification Resolver: Upon capture, in the identification variant, S selects an agent index guess g∈{1…n}. Probability of correct identification is parameterized as P(correct)=0.70+0.30×confidence. If correct (or in pure-tag variant), Ek becomes the new Seeker, ending the round with victory for S. If incorrect, Ek is released to a cell 2 units away, and play resumes. R8. [gap-fill] Round Time Limit: A round terminates after at most Tmax⁡=100 ticks. If no capture occurs by Tmax⁡, evaders win the round, and Seeker role rotates to a random evader. ## Edge cases - Tie at the end condition: Not applicable; capture is an atomic event occurring sequentially within tick resolution (first evader occupying the captured cell is tagged). - A player with no legal move: All agents always possess the `wait` (no movement) action. - Repetition / cycles: Bounded strictly by R8 (Tmax⁡=100 tick ceiling), preventing infinite chase loops. - Simultaneous claims on the same resource or position: Multiple evaders may occupy the same cell; if Seeker reaches that cell, the evader with the lowest index is captured first. - Player elimination mid-round: In multi-round elimination variants, captured evaders transition to spectator/graveyard state; remaining evaders continue until 1 survivor remains. - Resource exhaustion: Not applicable; spatial coordinates are bounded and persistent. - Termination guarantee: Provably guaranteed by the hard upper bound of Tmax⁡=100 ticks per round combined with atomic capture conditions. ## State representation - Persistent state: Arena bounds G, round counter, cumulative scores per agent, current active roles (`seeker_id`, `evader_ids`). - Per-tick ephemeral state: Coordinates (xS,yS) for Seeker; coordinates (xEi,yEi) for each Evader; pending action intents; acoustic sound wavefront emissions; tick timer t∈[1,Tmax⁡]. - Observable information: - Seeker observes: only noisy auditory directional vectors, bump collisions with arena walls, and successful capture contact. - Evaders observe: perfect public state (seeker position, other evader positions, arena bounds, tick counter). ## Action space - `move(dx, dy)`: Δx,Δy∈{−1,0,1} such that |Δx|+|Δy|≤1. - `tease(loudness)`: emit sound pulse with integer intensity ∈{0,1,2}. - `reach(dir)`: attempt capture sweep in specified cardinal direction or current cell. - `guess(agent_id)`: submitted upon capture trigger in identification mode. ## Playability notes - Single-player-vs-AI: Highly engaging in both configurations: (1) Human plays blindfolded Seeker using positional audio/radar cues against evasive heuristic AI, or (2) Human plays Evader sneaking around an AI Seeker implementing Bayesian particle filtering to estimate evader coordinates. - Group games (n>4): Reduces naturally to a 1v1 pursuit-evasion duel (n=1) on a smaller 7×7 grid without losing the fundamental sensory asymmetry. - Physical-skill games: Physical tactile recognition and auditory localization are formalized via noisy observation vectors and parameterized stochastic identification resolvers. - What is lost: The physical comedy of human blindfolded stumbling, comedic tactile misidentification of objects/furniture, and chaotic multi-person laughing banter are replaced by probabilistic POMDP inference. ## Implementation hints Treat the Seeker as an agent solving a Partially Observable Markov Decision Process (POMDP). Maintain a 2D probability occupancy grid updated via Bayesian updates each time an acoustic signal is emitted or an empty cell is reached. First milestone: implement 1v1 discrete grid chase with perfect auditory cues, then add distance-based Gaussian noise and mandatory tease timers.

Download as JSON

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

Download JSON
Cite this page

Blind Old Woman (Blind Man's Buff) (Czechia). (n.d.). Games of Countries. https://gamesofcountries.com/games/cz-czechia/slepa-baba

@misc{gamesofcountries-cz-czechia-slepa-baba,
  title = {Blind Old Woman (Blind Man's Buff) (Czechia)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/cz-czechia/slepa-baba}
}