Skip to main content
Formal Specification

Blind Goat — Formal Specification

Estonia · chase-tag · Confidence: high

← Back to game page

Implementation Suitability

Suitability: adaptation-requiredFidelity: discretizedGap-fills: 7
# Pimesikk (Blind Goat) — AI-optimized specification ## General idea A blindfolded acoustic tracking and tactile identification game played within an enclosed space. The blindfolded seeker ("blind goat") navigates solely via directional sound cues emitted by dodging evaders, attempting to corner, grab, and correctly deduce the real identity of a captive by tactile inspection while players enforce automated perimeter boundary safety warnings ("Tuli!"). ## Formal model - **Abstraction taken:** Complete visual occlusion is modeled by restricting the Seeker's information set to zero visual coordinates, replacing vision with an acoustic signal matrix (bearing and estimated distance from evader actions). Tactile recognition is modeled as a feature-matching deduction or Bayesian identification check. - **Player/agent count:** 1 Seeker agent (S), N Evader agents (E1…EN, N≥3). Total agents =N+1. - **Turn model:** Synchronous discrete ticks; each round terminates upon successful capture and correct identity verification. ## Formal rules R1. Setup and Boundary: - Enclosed 2D room of dimensions W×H (e.g. 10 m×10 m) [gap-fill: arena size standardized to 10m x 10m]. - Seeker starts at center (0,0) with status `BLINDFOLDED` (Vision Range =0). - Evaders start distributed randomly at distance d≥2.0 m from center. - Each agent is assigned a unique observable tactile feature vector 𝐟i∈ℝD (e.g. height, hair texture, sleeve cuff type) [gap-fill: tactile feature vector of dimension D=3]. R2. Movement & Acoustic Emission: - Each tick, agents submit simultaneous actions: - Evaders move up to vevader=1.0 m/tick and select noise emission level A∈{SILENT,WHISPER,CLAP,TAUNT}. - Higher noise levels generate public audio pings with bearing θi and distance noise σd∝1/Volume. - Seeker moves up to vseeker=0.8 m/tick in direction θS based on auditory estimates. R3. Safety Hazard Warning ("Tuli!"): - If the Seeker approaches within dhazard=0.5 m of a wall boundary, the engine automatically issues a public `TULI_WARNING` [gap-fill: boundary distance threshold set to 0.5m]. - On receiving `TULI_WARNING`, the Seeker's movement is deflected 180∘ away from the boundary without collision. R4. Grabbing & Freezing: - If the Euclidean distance between Seeker and Evader Ek satisfies d(S,Ek)≤Rreach=0.8 m, Seeker can execute `grab()`. - If successful, Ek enters status `HELD` and cannot move for Tinspect=2 ticks. R5. Tactile Identification: - During `HELD`, the Seeker observes a noisy sample of the captive's feature vector 𝐟ˆk∼N(𝐟k,σtouch2). - The Seeker submits `identify(suspect_id)`. - Outcome: - If `suspect_id` ==Ek: Identification succeeds. Ek becomes the new Seeker for the subsequent round. The round ends. - If `suspect_id` ≠Ek: Identification fails. Ek is immediately released and granted a 3-tick immunity sprint; Seeker continues searching. R6. [gap-fill: If the Seeker fails 5 consecutive identification guesses, or if round duration reaches Tround_max=150 ticks, the Seeker is granted a 1-tick sonar echo ping revealing the exact bearing of the nearest evader]. ## Edge cases - Tie at the end condition: Not applicable; exactly one evader is grabbed and verified per round transition. - A player with no legal move: Not applicable; agents always have valid movement or pass actions. - Repetition / cycles: Bounded by the sonar assist in R6 and finite round duration Tround_max [gap-fill: cycle bound via timeout and sonar assistance]. - Simultaneous claims on the same resource or position: If Seeker reaches range of multiple evaders, grab targets the closest agent; ties broken by lowest agent ID [gap-fill: grab target tiebreaker]. - Player elimination mid-round; turn order repair: Role transfer directly swaps the Seeker and captured Evader, preserving continuous execution. - Resource exhaustion: Not applicable; continuous acoustic signals and space. - Termination guarantee: Since vseeker>0, arena is bounded, and R6 provides sonar pings on stall, capture and correct identification occurs with probability 1 as t→∞ [gap-fill: termination guarantee via bounded space and perception assistance]. ## State representation - Persistent state: - `seeker_id`: Agent ID currently blindfolded. - `failed_guesses_count`: Integer count of failed identifications. - `round_number`: Integer r≥1. - Per-tick ephemeral state: - `positions`: Map of Agent ID →(x,y). - `audio_signals`: List of active acoustic pulses with source bearings and volumes. - `held_target`: Agent ID currently grabbed (or `None`). - `round_tick`: Integer t∈[1,Tround_max]. - Observable information: - Evaders observe: Full 2D ground truth positions of all agents and boundaries. - Seeker observes: Only active acoustic pulses, touch feedback when in contact, and `TULI_WARNING` signals. ## Action space - `move(dx, dy)`: Move within velocity limits. - `emit_sound(level)`: (Evaders) Choose `SILENT`, `WHISPER`, `CLAP`, or `TAUNT`. - `grab()`: (Seeker) Attempt grab on adjacent evader. - `identify(target_id)`: (Seeker) Submit guessed identity of captive. ## Playability notes - Single-player-vs-AI: Exceptional sensory puzzle game. Human plays Blind Seeker navigating via 3D spatialized audio pings, or human plays stealth Evader taunting AI Seeker. - Group games (n>4): Reduces to 1 Seeker vs 1 Evader (2 agents) cleanly, retaining the core cat-and-mouse acoustic dynamic. - Physical-skill games: Spatial hearing and tactile cloth/hair inspection are abstracted into noisy bearing estimators and feature vector classifiers. - What is lost in the adaptation: The physical darkness and disorientation of a real cloth blindfold, the giggling whispers in a farmhouse parlor, and the instinctive physical reach. ## Implementation hints Represent audio cues as Gaussian distributions over bearing angles. First milestone: Implement acoustic signal emission and gradient-following AI seeker before adding tactile feature vectors.

Download as JSON

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

Download JSON
Cite this page

Blind Goat (Estonia). (n.d.). Games of Countries. https://gamesofcountries.com/games/ee-estonia/pimesikk

@misc{gamesofcountries-ee-estonia-pimesikk,
  title = {Blind Goat (Estonia)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/ee-estonia/pimesikk}
}