Skip to main content
Formal Specification

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

Romania · guessing-hidden · Confidence: high

← Back to game page

Implementation Suitability

Suitability: engine-readyFidelity: discretizedGap-fills: 1
# Baba oarba — AI-optimized specification ## General idea An asymmetric, spatial pursuit and Bayesian identification game in a bounded 2D grid. A blindfolded seeker with zero visual range navigates via noisy auditory acoustic cues emitted by sighted agents. Upon trapping an agent in an adjacent cell, the seeker inspects a set of noisy tactile/vocal features to deduce the captive's identity and transfer the blindfold. ## Formal model - **Abstraction taken:** Bounded 2D grid (W×H, e.g. 10×10). Sighted agents emit acoustic ping signals; seeker navigation relies on gradient ascent over sound intensity. Captive identification is modeled as Bayesian feature matching over agent profile vectors (height, hair texture, disguised voice pitch). - **Player/agent count:** 1 Seeker (B) and K Sighted Agents (S1..SK, K∈[3..7], default K=4). - **Turn model:** Real-time ticked simultaneous execution. ## Formal rules R1. Setup: - Bounded grid W×H. - Seeker B starts at center (W/2,H/2) with visual field =∅. - Sighted agents S1..SK distributed randomly at distance d≥3 from seeker. - Each agent i has an invariant tactile profile vector 𝐯i=[hi,wi,pi]∈[0,1]3. R2. Movement & Audio Pings per tick: - Sighted agents choose move Δx,Δy∈{(−1,0),(1,0),(0,−1),(0,1),(0,0)} within boundaries. - Sighted agents emit auditory ping Ai(t)∈{0,1} (taunt / clap). - Seeker receives composite acoustic direction vector: 𝐃audio=∑i:Ai(t)=1𝐩i−𝐩B‖𝐩i−𝐩B‖2+𝝐(𝝐∼𝒩(𝟎,σ2𝐈)) - Seeker chooses move vector based on 𝐃audio. R3. Capture Event: - If Seeker B and a sighted agent Si occupy the same cell or adjacent orthogonal cells, B executes `grasp()`. - Agent Si is immobilized (`status = captured`). R4. Tactile Identification Phase: - Seeker receives noisy observation of captive profile: 𝐯ˆ=𝐯i+𝜼 (𝜼∼𝒩(𝟎,0.152𝐈)). - Seeker outputs guessed agent ID iˆ∈{1..K}. R5. Resolution & Role Transfer: - If iˆ==i (Correct Guess): Round terminates. Agent Si becomes the new Seeker B for the next round. - If iˆ≠i (Incorrect Guess): Captive Si is released (`status = free`), placed at distance d=3, and the seeker continues hunting. R6. [gap-fill] Mercy Timeout: If seeker fails 4 consecutive captures, the blindfold automatically transfers to the sighted agent with lowest lifetime evasion distance. ## Edge cases - Tie at the end condition: Not applicable — identification produces binary success/failure. - A player with no legal move: Waiting / moving 1 cell is always legal within grid boundaries. - Repetition / cycles: Bounded by R6 (maximum 4 failed identification cycles per round). - Simultaneous claims on the same resource or position: Sighted agents may share cells; if seeker enters a cell with multiple agents, seeker grasps the agent with smallest index. - Player elimination mid-round; turn order repair: Sighted agents are never eliminated; role transfer resets the single seeker. - Resource exhaustion: Not applicable. - Termination guarantee: Enforced by finite grid dimensions, acoustic convergence, and R6 mercy timeout. ## State representation - Persistent state: - `seeker_id`: Integer index 0..K. - `agent_profiles`: Matrix of K+1 vectors in [0,1]3. - `failed_guesses`: Integer counter. - `tick`: Integer counter. - Per-tick ephemeral state: - `positions`: Array of K+1 coordinates. - `audio_signals`: Array of K booleans. - Observable information: Sighted agents observe full grid. Seeker observes only noisy gradient vector 𝐃audio and profile vector 𝐯ˆ upon capture. ## Action space - Sighted Agent: `move(direction)`, `emit_ping()`. - Seeker: `move(direction)`, `grasp()`, `accuse_identity(agent_id)`. ## Playability notes - Single-player-vs-AI: Exceptional sensory-deprivation game. A human player as seeker must navigate via spatial audio cues and match tactile features against a known player roster. - Group games (n>4): Naturally supports K=3..10 sighted agents. - Physical-skill games: Physical blindfolded orientation and voice recognition converted into acoustic gradient descent and vector classification. - What is lost: Sensory touch of real clothing, laughter, and the fear of bumping into furniture. ## Implementation hints - Implement as a partially observable Markov decision process (POMDP). - Seeker policy: Kalman filter over sighted agent position distributions combined with maximum-likelihood profile matching.

Download as JSON

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

Download JSON
Cite this page

Blind Old Woman (Romanian Blind Man's Buff) (Romania). (n.d.). Games of Countries. https://gamesofcountries.com/games/ro-romania/baba-oarba

@misc{gamesofcountries-ro-romania-baba-oarba,
  title = {Blind Old Woman (Romanian Blind Man's Buff) (Romania)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/ro-romania/baba-oarba}
}