Skip to main content
Formal Specification

Blind Mice (Blind Man's Buff) — Formal Specification

Slovenia · chase-tag · Confidence: low

← Back to game page

Implementation Suitability

Suitability: engine-readyFidelity: discretizedGap-fills: 2
# Slepe miši — AI-optimized specification ## General idea An imperfect-information spatial pursuit game on a bounded 2D grid where a blindfolded seeker agent navigates exclusively via noisy acoustic sound-emission pings, captures nearby evader agents, and executes a Bayesian feature-matching guess to identify the captured agent. ## Formal model - **Abstraction taken:** The 2D playing arena is discretized into an L×W grid (8×8 cells). The seeker observes no visual coordinates (posseeker is private/unmapped to vision); evaders emit directional sound tokens ω∼Poisson(λ) per tick. Capture occurs on co-location, followed by a feature-identification test. - **Player/agent count:** N agents (Ablind,A1…AN−1, with N≥4). - **Turn model:** Synchronous discrete time ticks (Δt=1.0 s). ## Formal rules R1. Setup: - Bounded grid [0,7]×[0,7]. Evaders initialized at distinct random coordinates. - Seeker initialized at (3,3). Each evader possesses a distinct feature vector 𝐟i∈{0,1}D (representing height, clothing texture, hair length). R2. Evader Actions (per tick): - Action `move(dx, dy)` with ‖(dx,dy)‖≤1. - Action `emit_sound()`: Mandatory acoustic ping every Tping=3 ticks, broadcasting distance vector 𝐝=pos(Ai)−pos(Ablind)+ϵnoise to seeker. R3. Seeker Action (per tick): - Seeker executes `step(dx, dy)` with ‖(dx,dy)‖≤1 based on acoustic gradient estimation. R4. Capture and Identification: - If ‖pos(Ablind)−pos(Ak)‖≤0.5, evader Ak is captured. - Seeker samples K noisy tactile features 𝐟ˆ∼𝒩(𝐟k,σtactile2I). - Seeker submits identity guess kˆ∈{1,…,N−1}. - If kˆ==k: Guess is correct; Ak becomes the new Ablind for the next round. - If kˆ≠k: Guess fails; Ak is teleported 3 cells away, and search resumes. R5. Match Structure: - [gap-fill] Game runs for M=8 rounds. The agent who spent the fewest total rounds as Ablind wins. ## Edge cases - Tie at the end condition: Tied seeker durations result in a shared victory. - A player with no legal move: Inward bounce vectors enforce movement within grid bounds [0,7]2. - Repetition / cycles: [gap-fill] If seeker fails to capture any evader within Tmax=60 ticks, a forced capture of the closest evader is triggered. - Simultaneous claims on the same resource or position: Multiple evaders in the same cell are captured in order of proximity. - Player elimination mid-round; turn order repair: If an evader disconnects, their feature vector is removed from candidate set. - Resource exhaustion: Non-consumable spatial arena. - Termination guarantee: Finite round count M=8 and bounded tick timer Tmax guarantee finite termination. ## State representation - Persistent state: `blind_id: int`, `round_index: int`, `seeker_durations: dict[int, int]`. - Ephemeral tick state: `grid_positions: dict[int, tuple[int, int]]`, `sound_pings: list[tuple[float, float]]`, `captured_id: Optional[int]`. - Observable information: Evaders observe all positions; Seeker observes ONLY acoustic pings and tactile features upon contact. ## Action space - `move(dx: int, dy: int)`: Grid displacement. - `guess_identity(agent_id: int)`: Seeker classification action upon capture. ## Playability notes - Single-player-vs-AI: Highly engaging; human plays as seeker tracking acoustic audio pings and tactile features, or as evader dodging in silent zones. - Group games (n>4): Natively designed for 4 to 12 agents. - Physical-skill games: Acoustic localization modeled via Gaussian noise ϵnoise; tactile identification modeled via feature distance metrics. - What is lost: Physical fabric texture, genuine disorientation from spinning. ## Implementation hints Discrete 2D POMDP (Partially Observable Markov Decision Process). First milestone: implement grid movement and acoustic ping broadcasting before adding Bayesian feature classification.

Download as JSON

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

Download JSON
Cite this page

Blind Mice (Blind Man's Buff) (Slovenia). (n.d.). Games of Countries. https://gamesofcountries.com/games/si-slovenia/slepe-misi

@misc{gamesofcountries-si-slovenia-slepe-misi,
  title = {Blind Mice (Blind Man's Buff) (Slovenia)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/si-slovenia/slepe-misi}
}