Skip to main content
Formal Specification

The Lost One (Blindfold Touch & Identity Guessing) — Formal Specification

Belarus · guessing-hidden · Confidence: medium

← Back to game page

Implementation Suitability

Suitability: adaptation-requiredFidelity: probabilistic-proxyGap-fills: 1
# Prapavshchyk (Прапаўшчык) — AI-optimized specification ## General idea A traditional Belarusian parlor deduction game where a blindfolded central seeker ("The Lost One") examines silent challengers through tactile contact alone. Surrounding players may disguise wearable cues (swapping rings, hats, or sleeves); the seeker attempts to deduce the challenger's true identity within two guesses before forfeits or role transitions occur. ## Formal model - **Abstraction taken:** Each player possesses an immutable physical attribute vector (hand size, skin texture, height) and a mutable attire/accessory vector (rings, sleeve fabric, bracelets). Blindfold examination is modeled as noisy feature sampling from the challenger's active attribute vector. - **Player/agent count:** 1 Seeker agent (S), N Circle Player agents (C1,…,CN, with N∈[4,14]). - **Turn model:** Sequential challenge rounds. Each challenge consists of a disguise preparation step, tactile sampling queries by the Seeker, and up to two identity guesses. ## Formal rules R1. Setup: - One agent is designated as Seeker S. - Each circle player Ci has a ground-truth profile of base features Fbase(Ci) and mutable attire features Fattire(Ci). - Seeker maintains a belief distribution over all circle players. R2. Disguise Step: - Before a challenger steps forward, circle players may execute `swap_item(item_id, target_player)` to trade attire features. - The Game Engine (Leader) selects challenger C∗∈{C1,…,CN} uniform-randomly without revealing identity to S. R3. Tactile Inspection Phase: - Seeker S selects up to ksample inspection actions (e.g. `inspect_palms()`, `inspect_wrist()`, `inspect_sleeve()`). - Engine returns noisy observation vector O∼𝒩(Factive(C∗),σ2). R4. Guessing & Resolution: - Seeker submits `guess_identity(predicted_id)`. - **Outcome A (Correct):** If predicted_id=C∗, the guess is correct. Challenger C∗ becomes the new Seeker for the next round; S joins the circle. - **Outcome B (Incorrect 1st Attempt):** Engine returns `INCORRECT`. Seeker may submit a 2nd guess `guess_identity(predicted_id_2)`. - **Outcome C (Incorrect 2nd Attempt):** Challenger C∗ returns to the circle unmasked. Seeker increments `consecutive_failures` by 1. R5. Forfeit & Reset: - If `consecutive_failures` reaches 3, Seeker executes `pay_forfeit()` and `consecutive_failures` resets to 0 [gap-fill: forfeit modeled as a score penalty]. - Engine selects a new challenger from the remaining untried circle players. R6. Round Termination: A round terminates the instant a challenger is correctly identified. ## Edge cases - Tie at the end condition: Not applicable; individual deduction rounds resolve decisively into identification or failure. - A player with no legal move: Not applicable; Seeker must submit queries and guesses; challengers must submit to inspection. - Repetition / cycles: Challengers are drawn without immediate replacement from the active circle until identified. - Simultaneous claims on the same position: Strictly one challenger approaches the Seeker at a time. - Elimination mid-round and turn-order repair: Circle players are not eliminated; role swap transfers Seeker status cleanly. - Resource exhaustion: Inspection samples ksample and guesses per candidate (2) are finite bounds. - Termination guarantee: Each challenger receives at most 2 guesses. With finite players N and a non-zero probability of correct identification from unmasked physical traits, the probability of identifying a player approaches 1 over repeated challenges. ## State representation - Persistent state: - `seeker_id`: Agent ID of current seeker. - `consecutive_failures`: Integer [0,3]. - `player_features`: Map of player ID → feature vector. - Per-challenge ephemeral state: - `active_challenger_id`: Hidden agent ID C∗. - `inspection_count`: Integer [0,ksample]. - `guesses_used`: Integer [0,2]. - `seeker_observations`: List of observed feature values. - Observable information: Seeker observes only noisy sampled features and guess outcome feedback; Circle players observe disguise trades and public round outcomes. ## Action space - `swap_item(item_type, target_player)`: Circle player attire swap. - `inspect_palms()`, `inspect_wrist()`, `inspect_sleeve()`: Seeker tactile sampling queries. - `guess_identity(player_id)`: Seeker formal guess. - `pay_forfeit()`: Seeker penalty resolution after 3 failed candidates. ## Playability notes - Single-player-vs-AI: Natural Bayesian deduction puzzle where human Seeker gathers tactile clues against AI opponents who strategically trade clothing/accessories. - Group reduction: Reducible to 1 Seeker and 3 Circle players while retaining deduction complexity. - Physical-skill games: Tactile skin sensation, thermal cues, and subtle hand callouses are converted into feature vector values and noise variances. - What is lost: The intimate social atmosphere of the candlelit cottage, subtle breathing sounds, and repressed giggles. ## Implementation hints Model as a classic Bayesian classification problem over discrete player categories with binary/continuous feature vectors. First milestone: implement feature observation generation and 2-guess state verification before adding the attire-swapping mechanic.

Download as JSON

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

Download JSON
Cite this page

The Lost One (Blindfold Touch & Identity Guessing) (Belarus). (n.d.). Games of Countries. https://gamesofcountries.com/games/by-belarus/prapavshchyk

@misc{gamesofcountries-by-belarus-prapavshchyk,
  title = {The Lost One (Blindfold Touch & Identity Guessing) (Belarus)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/by-belarus/prapavshchyk}
}