Formal Specification
Blind Goat / Portuguese Blind Man's Buff — Formal Specification
Portugal · guessing-hidden · Confidence: low
← Back to game pageImplementation Suitability
Suitability: engine-readyFidelity: probabilistic-proxyGap-fills: 4
# Jogo da Cabra-Cega — AI-optimized specification
## General idea
An asymmetric pursuit-evasion game with hidden information. A blindfolded seeker navigates an 8×8 grid using noisy acoustic observations to corner evasive agents. Upon contact, the seeker attempts to deduce the captured agent's identity based on feature profiles, with role swaps occurring on successful identification.
## Formal model
- **Abstraction taken:** Continuous physical space is mapped to an 8×8 discrete grid. The blindfold condition is modeled by restricting the seeker's observation to noisy acoustic signals (distance + direction with Gaussian noise) while evasive agents have full local grid visibility.
- **Player/agent count:** 4 agents (Agent 0 = Seeker; Agents 1..3 = Evasive Runners).
- **Turn model:** Simultaneous movement ticks (`dt = 1` turn).
## Formal rules
R1. Setup: 8×8 grid `(x, y) in {0..7}^2`. Agent 0 (Seeker) starts at `(3, 3)`. Agents 1..3 start at random perimeter cells. Each evasive agent has a discrete profile `(Height: S/M/L, Hair: Short/Long, Texture: Smooth/Rough)` [gap-fill].
R2. Movement Phase (per tick):
- Seeker chooses a step in `{(0,1), (0,-1), (1,0), (-1,0), (0,0)}`.
- Evasive agents simultaneously choose a step in the same orthogonal set.
- Boundaries: Moving outside `{0..7}^2` is illegal.
R3. Observation Model:
- Seeker receives noisy acoustic pings: For each evasive agent `i`, Seeker observes `d_noisy(i) = d_true(i) + N(0, 1.0)` and an approximate quadrant [gap-fill].
- Evasive agents observe Seeker's exact coordinate.
R4. Contact Check:
- If Seeker and any Evasive Agent `k` occupy the same cell, a Capture Event occurs.
R5. Tactile Identification Phase:
- Seeker queries 2 physical features of the captured agent (e.g. `Query(Height)`, `Query(Hair)`).
- Seeker makes an identity guess `Guess in {Agent 1, Agent 2, Agent 3}`.
- If `Guess == k`: Correct! Role swap occurs (Agent `k` becomes the new Seeker, Agent 0 becomes an evasive runner). Round ends; Seeker receives 1 point.
- If `Guess != k`: Incorrect! Captured agent is pushed to a random adjacent cell; pursuit resumes [gap-fill].
R6. Win condition: Match terminates after 6 successful role transitions. The agent with the most identification points wins.
## Edge cases
- Tie at the end condition: If agents tie in points after 6 transitions, a 2-transition playoff is conducted.
- A player with no legal move: Not applicable; `(0,0)` (staying in place) is always legal.
- Repetition / cycles: Bounded by max 100 ticks per hunt; if no capture occurs within 100 ticks, Seeker is rotated [gap-fill].
- Simultaneous claims on the same resource or position: Evasive agents may co-occupy cells without collision penalty.
- Player elimination mid-round: If an evasive agent disconnects, they are removed from the candidate pool; if Seeker disconnects, the next agent in order becomes Seeker.
- Resource exhaustion (pieces, tokens, space, time): Tick cap per hunt ensures finite game duration.
- Termination guarantee: The 100-tick cap per hunt and 6-transition match limit guarantee finite termination.
## State representation
- Persistent state: `points: [int; 4]`, `current_seeker: int`, `transition_count: int (0..6)`.
- Per-tick ephemeral state: `agent_positions: [(int, int); 4]`, `tick_counter: int`, `feature_profiles: [Profile; 3]`.
- Information sets:
- Seeker observes: own position, noisy acoustic pings, queried features upon capture.
- Evasive agents observe: all agent positions.
## Action space
- Movement: `move(dx: int, dy: int)` where `dx, dy in {-1, 0, 1}` with `|dx| + |dy| <= 1`.
- Tactile Guess (Seeker only, upon capture): `guess_identity(target_agent_id: int)`.
## Playability notes
- Single-player-vs-AI: Engine-ready. Player controls Seeker (deductive hunt) or an Evasive Agent (stealth evasion).
- Group games (n>4): Native 4-agent multi-agent environment with asymmetric information.
- Physical-skill games: Tactile feeling and auditory detection modeled as Bayesian belief updating over agent identity and position.
- What is lost in the adaptation: Physical touch sensation, laughing, and erratic physical dodging.
## Implementation hints
Discrete 8×8 grid with hidden agent attributes. Suggested milestone 1: Implement the 2D grid movement, noisy observation generator, and capture/identification resolution logic.
Download as JSON
Download the full game corpus entry for this game as structured JSON.
Download JSONCite this page
Blind Goat / Portuguese Blind Man's Buff (Portugal). (n.d.). Games of Countries. https://gamesofcountries.com/games/pt-portugal/jogo-da-cabra-cega
@misc{gamesofcountries-pt-portugal-jogo-da-cabra-cega,
title = {Blind Goat / Portuguese Blind Man's Buff (Portugal)},
year = {n.d.},
publisher = {Games of Countries},
url = {https://gamesofcountries.com/games/pt-portugal/jogo-da-cabra-cega}
}