Skip to main content
Formal Specification

The Ring (Guernsey Ring Game) — Formal Specification

Guernsey · guessing-hidden · Confidence: medium

← Back to game page

Implementation Suitability

Suitability: engine-readyFidelity: directGap-fills: 3
# La Bague (Guernsey) — AI-optimized specification ## General idea A hidden-token tracking and deception game on a circular graph. A single ring token travels secretly among N−1 circle agents connected along a string loop. Circle agents execute noisy transfer/bluff actions while a central Seeker agent filters noisy observations to locate the token holder within a limited budget of accusations. ## Formal model - **Abstraction taken:** The continuous passing along the string is modeled as synchronous discrete ticks t=1,2,…. At each tick, the token holder chooses to hold or pass the token to an adjacent node on the ring graph CN−1. - **Player/agent count:** N agents (Agent 0: `SEEKER`, Agents 1…N−1: `CIRCLE_PLAYERS`). - **Turn model:** Two-step turn cycle per tick: 1. Circle Phase: Simultaneous secret movement/bluff action by all circle agents. 2. Seeker Phase: Seeker inspects observation vector 𝐎t and either passes or makes an accusation. ## Formal rules R1. Setup: - Circle agents 1,…,M (where M=N−1) form a ring graph CM with adjacency i↔(i±1(modM)). - The token `RING` is assigned randomly: posring∼𝒰({1,…,M}). - Seeker is initialized with guess budget G=3. R2. Circle Secret Action: - Let k=posring. Agent k selects action ak∈{HOLD,PASS_CW,PASS_CCW}. - Token position updates to posring←Neighbor(k,ak). - Agents j≠k select bluffing actions. R3. Observation Emission: - Each agent i∈{1,…,M} emits a continuous signal si∈ℝ reflecting physical motion noise: si∼𝒩(1.0,σ2) if agent i was involved in a pass; si∼𝒩(0.0,σ2) otherwise. - Seeker receives public vector 𝐒t=(s1,…,sM). R4. Seeker Decision: - Seeker chooses either: - `wait()`: Advance to tick t+1. - `challenge(target_id)`: Accuse circle agent target_id∈{1,…,M}. R5. Challenge Evaluation: - If target_id==posring: Accusation succeeds. Seeker gains +1 point. The accused agent is assigned as the new `SEEKER`, and the previous Seeker enters the circle at the accused agent's vacant node. Round ends. - If target_id≠posring: Accusation fails. Guess budget G←G−1. - If G==0, the Seeker fails the round; circle team scores +1 point, and a new round begins with the same Seeker [gap-fill: failed seeker repeats round]. R6. [gap-fill] Match Horizon: Match terminates after a fixed series of R=8 rounds. The agent with the highest individual score as Seeker wins. ## Edge cases - Tie at the end condition: Tied agents at round R=8 are ranked by lowest average ticks per successful capture. - A player with no legal move: Not applicable; circle agents always have valid transfer choices, and the Seeker always has `wait` or `challenge`. - Repetition / cycles (and the loop-breaking rule): [gap-fill] If Seeker takes no action within Tmax=25 ticks, the round automatically concludes as a failed seeker attempt. - Simultaneous claims on the same resource or position: Not applicable; exactly one token exists and accusations are atomic. - Player elimination mid-round; turn order repair: No player elimination; roles rotate dynamically upon successful challenge. - Resource exhaustion (pieces, tokens, space, time): Single token; guess count strictly bounded (G≤3); ticks bounded (T≤25). - Termination guarantee: Guaranteed by round cap R=8 and per-round tick bound Tmax=25. ## State representation - Persistent state: - `round`: Integer ∈[1,8]. - `seeker_id`: Agent index ∈{0,…,N−1}. - `scores`: Array of size N. - Ephemeral state (per tick): - `tick`: Integer ∈[1,25]. - `token_holder`: Node index ∈{1,…,M}. - `guesses_left`: Integer ∈[0,3]. - `signal_history`: Sequence of signal vectors. - Observable information: - Circle agents observe their own position and whether they hold the token. - Seeker observes only the public signal history, guesses remaining, and node indices. ## Action space - Circle Agent i: `action` ∈{HOLD,CW,CCW}. - Seeker: `wait()` or `challenge(node_id)`. ## Playability notes - Single-player-vs-AI: Highly playable; tests human probabilistic deduction against simulated bluffers, or human bluffing timing against an optimal Bayesian seeker. - Group games (n>4): Scales naturally for any N∈[4,18]. - Physical-skill games: Abstracted as Gaussian signal discrimination with parameter σ. - What is lost in the adaptation: Subtle micro-gestures, vocal cadence changes in the song, and physical contact cues. ## Implementation hints - Represent belief state as a probability distribution over the M circle nodes updated via standard HMM forward filter. - AI Seeker triggers `challenge(i)` when maxi⁡bt(i)>θthreshold. - First milestone: Validate uniform distribution over ring graph under random walk before testing Bayesian seeker accuracy.

Download as JSON

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

Download JSON
Cite this page

The Ring (Guernsey Ring Game) (Guernsey). (n.d.). Games of Countries. https://gamesofcountries.com/games/gg-guernsey/la-bague-guernesiaise

@misc{gamesofcountries-gg-guernsey-la-bague-guernesiaise,
  title = {The Ring (Guernsey Ring Game) (Guernsey)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/gg-guernsey/la-bague-guernesiaise}
}