Formal Specification
Blindfolded Goat Catching / Blind Man's Bluff — Formal Specification
Vietnam · guessing-hidden · Confidence: medium
← Back to game pageImplementation Suitability
Suitability: engine-readyFidelity: discretizedGap-fills: 5
# Bịt mắt bắt dê — AI-optimized specification
## General idea
Bịt mắt bắt dê is a classic Vietnamese sensory tag and identification game where a blindfolded seeker navigates inside a circular arena among evasive "goats." The goats periodically emit vocal bleats ("be be") giving noisy positional clues; the seeker uses auditory tracking to corner and tag a player, then performs a tactile inspection to deduce and announce the player's true identity. A correct identification transfers the catcher role.
## Formal model
- **Abstraction taken:** The physical space is modeled as a discretized 2D grid bounded within a circular radius . The blindfold is modeled as zero direct visual observation () with noisy acoustic vector observations from goat bleats. The tactile inspection upon tagging is modeled as a sampled feature vector (e.g., height, texture, clothing signature) with partial noise [gap-fill: spatial movement is discretized onto a 2D grid bounded by radius R with unit cell movement per tick].
- **Player/agent count:** 1 Catcher agent () and Goat agents (), where .
- **Turn model:** Simultaneous tick-based discrete movement per time step .
## Formal rules
R1. Arena and Setup:
- Arena is a discrete grid .
- Catcher starts at origin with state .
- goats start at distinct non-origin coordinates .
- Each goat has a static hidden identity attribute vector .
R2. Movement Phase (per tick ):
- Every agent simultaneously submits a movement action .
- New positions are clamped to remain strictly within .
- Goats choose whether to emit an acoustic cue . At least one goat must bleat every ticks [gap-fill: if no goat bleats for 3 consecutive ticks, a random goat is forced to bleat].
R3. Observation and Tagging:
- The Catcher receives an acoustic observation: a set of noisy bearings and distances to all active bleating goats.
- If at the end of tick , (Catcher shares cell with Goat ), a `TAG` event occurs. The tagged goat is frozen.
R4. [gap-fill] Guess Limit per Tag:
- Upon a `TAG` event, the Catcher is granted exactly 1 guess action `guess_identity(target_agent_id)`.
R5. [gap-fill] Tactile Observation Feature Vector:
- When a `TAG` occurs, the Catcher receives an observation , where .
- The Catcher submits `guess_identity(g)`.
- If (correct guess), the Catcher wins the round; Goat becomes the new Catcher for the next round, and the round terminates.
- If (incorrect guess), Goat is released, teleported to a random free cell at distance from the Catcher, and normal play resumes.
R6. [gap-fill] Step Limit / Turn Timeout:
- If ticks elapse without a successful identification, the round terminates with a victory for the Goats; the Catcher remains Catcher for a restart.
## Edge cases
- Tie at the end condition: Not applicable; victory occurs on the first correct identification event or on timeout expiration.
- A player with no legal move: All agents always have at least the `STAY` action available.
- Repetition / cycles: Bounded by the hard step limit in R6.
- Simultaneous claims on the same resource or position: If Catcher occupies a cell containing multiple goats simultaneously, the Catcher tags the goat with the lowest index .
- Player elimination mid-round; turn order repair: No players are eliminated; captured goats are merely frozen during inspection, then released or converted to catcher.
- Resource exhaustion: Not applicable; no consumable resources except time ticks bounded by .
- Termination guarantee: why the game provably ends: Guaranteed by R6 ().
## State representation
- Persistent state:
- `arena_radius`: integer .
- `catcher_pos`: coordinate .
- `goat_positions`: array of coordinates .
- `goat_attributes`: array of static feature vectors .
- `current_tick`: integer .
- `tagged_target`: nullable integer index of tagged goat.
- Per-round ephemeral state:
- `bleat_signals`: list of emitted sound coordinates in current tick.
- `tactile_sample`: noisy feature vector upon active tag.
- Observable information:
- Catcher: Sees NO visual positions of any agent (). Observes own position, wall collisions, acoustic bearing vectors from bleats, and upon tag.
- Goats: Fully observe catcher position, own position, and peer positions.
## Action space
- Catcher:
- `move(direction)`: direction .
- `guess_identity(agent_id)`: valid only during a `TAG` state.
- Goats:
- `move(direction)`: direction .
- `vocalize(bleat_flag)`: boolean flag to emit sound.
## Playability notes
- Single-player-vs-AI: Natural; human player can play as the blindfolded Catcher (using audio/tactile clues) against AI goats, or as an evasive goat against an AI Catcher using Bayesian tracking.
- Group games (n>4): The standard form naturally accommodates goats. To reduce to 2 agents (1 Catcher vs 1 Goat), the game operates identically to a pursuit-evasion duel inside the circle with identification based on a multi-feature quiz.
- Physical-skill games: Physical balance and blindfold navigation are abstracted into POMDP state estimation and acoustic triangulation.
- What is lost in the adaptation: The physical comedy of stumbling, tactile groping of clothing/hair, and laughter of live participants.
## Implementation hints
- Model as a standard POMDP / GridWorld environment (e.g., grid with circular mask).
- Set ticks, radius .
- Suggested first milestone: Implement the 2D grid kinematics with 1 Catcher and 4 Goats, verifying collision detection and the `guess_identity` trigger.
Download as JSON
Download the full game corpus entry for this game as structured JSON.
Download JSONCite this page
Blindfolded Goat Catching / Blind Man's Bluff (Vietnam). (n.d.). Games of Countries. https://gamesofcountries.com/games/vn-vietnam/bit-mat-bat-de
@misc{gamesofcountries-vn-vietnam-bit-mat-bat-de,
title = {Blindfolded Goat Catching / Blind Man's Bluff (Vietnam)},
year = {n.d.},
publisher = {Games of Countries},
url = {https://gamesofcountries.com/games/vn-vietnam/bit-mat-bat-de}
}