Formal Specification
Blind Old Woman (Blind Man's Buff) — Formal Specification
Slovakia · chase-tag · Confidence: low
← Back to game pageImplementation Suitability
Suitability: engine-readyFidelity: discretizedGap-fills: 4
# Slepá baba — AI-optimized specification
## General idea
A discretized asymmetric chase-and-tag game on a bounded 2D grid where one blindfolded chaser navigates exclusively via noisy directional sound cues emitted by evaders (based on the chosen mode: silent, clapping, or stomping), culminating in a feature-based tactile identity classification step upon capture.
## Formal model
- **Abstraction taken:** The playing arena is represented as a discrete toroidal/walled grid [gap-fill]. The blindfold effect is modeled by completely withholding evader grid coordinates from the Baba; the Baba receives only noisy acoustic vectors .
- **Player/agent count:** 1 Baba vs Evaders ().
- **Turn model:** Synchronous tick. Each tick, Baba moves up to 1 cell; Evaders move up to 1 cell.
## Formal rules
R1. Setup: Arena is a grid. Baba spawns at . Evaders spawn at random perimeter coordinates.
R2. Mode declaration: Baba selects an acoustic mode:
- Mode 1 (Silent): Evaders emit noise with probability .
- Mode 2 (Clapping): Evaders emit noise with probability , noise variance .
- Mode 3 (Tugging): Evaders must spend at least 1 turn adjacent to Baba every 5 turns [gap-fill]; high risk, high mobility.
R3. Movement tick: All agents simultaneously submit a movement vector .
R4. Acoustic observation: At the start of Baba's turn, if an evader emits noise, Baba receives an approximate direction vector .
R5. Tagging: If Baba moves into the same cell as an Evader , capture occurs immediately.
R6. Identification phase: Baba receives a feature vector of the captured agent (e.g. `[height, clothing_texture, voice_pitch]` with added perceptual noise). Baba must output an agent index guess .
- If (correct identification): The round ends; captured agent becomes the new Baba.
- If (incorrect guess): The captured agent is teleported 3 cells away in a random direction [gap-fill]; the hunt continues.
R7. Round limit: A round terminates after a maximum of 50 ticks [gap-fill]. If no correct capture occurs, Baba receives 0 score and a new round begins with the same Baba.
## Edge cases
- Tie at the end condition: Not applicable; identification produces a single binary success/failure.
- A player with no legal move: In any position, (staying in place) is always legal.
- Repetition / cycles: Bounded strictly by the 50-tick round cap.
- Simultaneous claims on the same resource or position: Evaders can share cells; if Baba enters a cell containing multiple evaders, Baba captures the first resolved evader index.
- Player elimination mid-round; turn order repair: No eliminations occur during active chasing; the game resets on successful role transition.
- Resource exhaustion (pieces, tokens, space, time): Grid boundaries constrain spatial movement; tick limit guarantees temporal exhaustion.
- Termination guarantee: The maximum tick limit (R7) guarantees deterministic termination of every round.
## State representation
- Persistent state: `round_number: int`, `current_baba_id: int`, `scores: dict[int, int]`, `tick: int`.
- Per-tick ephemeral state: `baba_pos: tuple[int, int]`, `evader_positions: dict[int, tuple[int, int]]`, `acoustic_signals: list[tuple[float, float]]`.
- Observable information: Evaders observe the full grid (Baba position and all peer positions). Baba observes ONLY their own position, boundary collisions, and noisy acoustic vectors received in R4.
## Action space
- `set_mode(mode_id: int)`: Select game mode .
- `move(direction: int)`: Move .
- `guess_identity(agent_id: int)`: Submit hypothesis of captured agent's true ID.
## Playability notes
- Single-player-vs-AI: Highly engaging either playing as Baba (deducing positions from noisy audio pings and tactile features) or as an Evader (stealth maneuvering).
- Group games (n>4): Naturally supports 4 to 15 agents; evaders make independent tactical evasion choices.
- Physical-skill games: Tactile and audio perception are modeled via Gaussian noise vectors over discrete feature attributes.
- What is lost in the adaptation: Physical room obstacles, live vocal banter, and direct touch sensation.
## Implementation hints
Low complexity. The core engine is a standard grid-world POMDP (Partially Observable Markov Decision Process) for the Baba agent. Implement the 2D grid movement first, then plug in the acoustic signal generator and feature guessing step.
Download as JSON
Download the full game corpus entry for this game as structured JSON.
Download JSONCite this page
Blind Old Woman (Blind Man's Buff) (Slovakia). (n.d.). Games of Countries. https://gamesofcountries.com/games/sk-slovakia/slepa-baba
@misc{gamesofcountries-sk-slovakia-slepa-baba,
title = {Blind Old Woman (Blind Man's Buff) (Slovakia)},
year = {n.d.},
publisher = {Games of Countries},
url = {https://gamesofcountries.com/games/sk-slovakia/slepa-baba}
}