Skip to main content
Formal Specification

Queenie-I-O / Queenie, Who Has the Ball? — Formal Specification

Ireland · guessing-hidden · Confidence: low

← Back to game page

Implementation Suitability

Suitability: adaptation-requiredFidelity: directGap-fills: 1
# Cluiche na Banríona — AI-optimized specification ## General idea An asymmetric game of imperfect information, bluffing, and deduction where a central "Queenie" tosses a ball backward over their head to a line of concealing players, and must identify which agent holds the ball behind their back using Bayesian inference over observed leakage signals against deceptive bluffing behaviors. ## Formal model - **Abstraction taken:** The ball throw and catch randomly selects one Line agent as the true `Ball_Holder`. Physical posture cues, micro-expressions, and nervousness are abstracted into a 1D scalar observation signal `S_i ∈ [0.0..1.0]` emitted by each Line agent `i`, where `S_i ~ N(μ_hold, σ²)` if holding the ball, or `S_i ~ N(μ_bluff, σ²)` if empty-handed. - **Player/agent count:** 5 agents total: 1 `Queenie` agent and 4 `Line` agents (`L_1, L_2, L_3, L_4`). - **Turn model:** Round-based sequential phase progression (Toss -> Catch/Conceal -> Chant/Signal Emission -> Deduction Phase) over a fixed series of 10 rounds. ## Formal rules R1. Game state initialization: Queenie role assigned to `A_0`. Line agents are `L = {A_1, A_2, A_3, A_4}`. Round index `round = 1..10`. Scores initialized to 0. R2. Toss & Concealment phase: - Queenie executes `toss_ball_blind()`. - Ball catcher `H ∈ L` is chosen uniformly at random: `P(H = L_i) = 0.25`. - `H` transitions to state `True_Holder`. Other line agents transition to state `Empty_Bluffer`. R3. Signal generation: - Each Line agent `L_i` declares a bluffing strategy `bluff_profile ∈ {poker_face, misdirection, nervous_feign}`. - Observable leakage signal `S_i` is generated: * If `L_i == H`: `S_i ~ N(0.70 - 0.20 * bluff_skill_i, 0.15)`. * If `L_i != H`: `S_i ~ N(0.30 + 0.15 * deception_bias_i, 0.15)`. - Queenie observes the vector of signals `[S_1, S_2, S_3, S_4]`. R4. Guessing phase: - Queenie executes up to 2 distinct guess queries: `guess_holder(target_agent_id)`. - If `target_agent_id == H`: * Guess is `CORRECT`. Queenie scores +10 points. * `H` becomes the new Queenie for the subsequent round; former Queenie joins the Line. - If `target_agent_id != H`: * Queenie executes second guess `guess_holder(target_2)`. * If second guess is correct: Queenie scores +5 points; `H` becomes new Queenie. * If second guess is wrong: Round ends in `FAIL`. True holder `H` scores +10 points for successful deception. Queenie retains the Queenie role. R5. Round progression: - `round` increments by 1. Roles and scores update. R6. Victory condition: - After 10 completed rounds, the agent with `argmax(score[agent_id])` wins. - [gap-fill] Tie-breaker: If scores are tied at round 10, the agent with the higher percentage of successful deductions wins. ## Edge cases - Tie at the end condition: Broken by deduction accuracy percentage; if identical, shared victory. - A player with no legal move: Not applicable; Queenie always selects valid agent indices from the Line set `L`. - Repetition / cycles: Bounded strictly by the 10-round match limit. - Simultaneous claims on the same resource or position: Guesses and signals are processed sequentially. - Player elimination mid-round: If a line agent drops out, the ball selection distribution renormalizes over remaining active line agents. - Resource exhaustion: Ball and rounds are strictly conserved. - Termination guarantee: Monotonically increasing round counter capped at 10 rounds guarantees termination. ## State representation - Persistent state: Cumulative scores `score[agent_id]`, current Queenie ID, round counter (1..10). - Per-player state: Skill ratings (`bluff_skill`, `deception_bias`, `deduction_skill`), historical guess success rate. - Per-turn ephemeral state: True ball holder `H`, emitted signal vector `[S_1, S_2, S_3, S_4]`, active guess history for current round. - Observable information: Queenie observes only the noisy signal vector `S` and past historical scores; true holder ID is hidden from Queenie until the round resolves. ## Action space - `toss_ball_blind()`: Queenie initiates round. - `set_bluff_profile(profile_type)`: Line agent configures behavioral posture. - `guess_holder(target_agent_id)`: Queenie commits deduction against specific line agent. ## Playability notes - Single-player-vs-AI: Excellent psychological deduction puzzle where player uses Bayesian filtering to identify the liar among AI agents. - Group games (n>4): Standard 5-agent configuration maps directly into multi-agent reinforcement learning or game-theoretic benchmark frameworks. - Physical-skill games: Physical body language and facial micro-expressions abstracted into continuous stochastic observation vectors. - What is lost: The contagious laughter and energetic chanting of children standing in a schoolyard line. ## Implementation hints Model Queenie's deduction as a Bayesian posterior update over hypothesis set `{H = L_1, ..., H = L_4}` given observed likelihoods `P(S | H = L_i)`.

Download as JSON

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

Download JSON
Cite this page

Queenie-I-O / Queenie, Who Has the Ball? (Ireland). (n.d.). Games of Countries. https://gamesofcountries.com/games/ie-ireland/cluiche-na-banriona

@misc{gamesofcountries-ie-ireland-cluiche-na-banriona,
  title = {Queenie-I-O / Queenie, Who Has the Ball? (Ireland)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/ie-ireland/cluiche-na-banriona}
}