Skip to main content
Formal Specification

String Figures / Cat's Cradle — Formal Specification

Japan · string-figure · Confidence: low

← Back to game page

Implementation Suitability

Suitability: adaptation-requiredFidelity: probabilistic-proxyGap-fills: 3
# Ayatori — AI-optimized specification ## General idea A two-player turn-taking topological transformation game played with a single closed loop of string. Players alternate taking over the string formation from their opponent's hands by pinching specific crossings and lifting underneath parallel strands to transition the structure into a new recognized geometric figure without tangling or collapsing the loop. ## Formal model - **Abstraction taken:** The continuous spatial physics of manual finger dexterity, string friction, and delicate loop extraction are abstracted into a directed state graph of recognized traditional figure topologies. The execution of a transition is governed by a parameterized success probability function `P(success) = f(difficulty, dexterity_skill)`. - **Player/agent count:** 2 players (Player 0 and Player 1). - **Turn model:** Sequential alternating turns. ## Formal rules R1. State space: The system exists in one of a finite set of recognized shape states: `S = {Tanbo, Kawa, Fune, Tsuzumi, Daiya, Ushi-no-Kubi, Kakemono, Fukurō, Failed}`. R2. Initial state: Player 0 opens the game by constructing `Tanbo` (Paddy Field) on their hands. R3. On a turn, the active player must choose one legal transformation edge outgoing from the current state `s_current` to a target state `s_next`. R4. Transition resolution: - With probability `P(success) = clamp(skill_active - difficulty(edge), 0.05, 0.95)`, the transition succeeds: the string formation transfers to the active player's hands, and the state becomes `s_next`. - With probability `1 - P(success)`, execution fails: the string tangles or collapses, transitioning to state `Failed`. R5. Turn transfer: If the transition succeeds, the active player displays `s_next` and becomes the holder; the turn passes to the other player. R6. Win/Loss condition: If a player causes the state to enter `Failed`, that active player immediately loses, and the opposing player wins. R7. [gap-fill] Cycle limit: If 50 consecutive successful transitions occur without a failure, the match is declared a draw. R8. [gap-fill] Difficulty scaling: Each successive loop cycle through the state graph increases edge difficulty by a factor of 1.05. R9. [gap-fill] Time limit per turn: An agent must declare its chosen transformation edge within 30 seconds, or forfeit the game. ## Edge cases - **Tie at the end condition:** Possible only via R7 (reaching the 50-transition cap without error). - **A player with no legal move:** If a shape state has no outgoing valid transitions (a terminal dead-end configuration), the player who produced that dead end is deemed to have failed, conceding the round. - **Repetition / cycles:** The state graph contains natural cycles (e.g., `Tanbo -> Kawa -> Fune -> Tsuzumi -> Tanbo`); bounded by R7 (50-move cap) and R8 (difficulty inflation). - **Simultaneous claims on the same resource or position:** Not applicable; turns and transfers are strictly sequential. - **Player elimination mid-round:** Not applicable; strictly a 2-player duel terminating on the first failed transfer. - **Resource exhaustion:** Only 1 string loop is used; never consumed or split. - **Termination guarantee:** Guaranteed by stochastic failure probability on every transition (`P(failure) >= 0.05`) and the hard cap of 50 turns (R7). ## State representation - **Persistent state:** `current_shape` in `S`, `turn_count` (int), `cycle_count` (int), `holder_player_id` (0 or 1). - **Per-player state:** `dexterity_skill` (float in `[0.0, 1.0]`). - **Observable information:** Fully observable perfect-information game. ## Action space - `select_transformation(target_shape)`: Choose which outgoing graph transition to attempt from the current configuration. ## Playability notes - **Single-player-vs-AI:** Highly natural; AI can use simple heuristic search balancing high-probability safe shapes against high-difficulty traps that force opponent mistakes. - **Group games (n>4):** Reduces to 2 active players at a time; extra players sit in rotation queue as challengers. - **Physical-skill games:** Physical finger-snatching is modeled via `P(success) = f(difficulty, skill)`. - **What is lost in the adaptation:** The sensory tactile experience of string tension, finger flexibility, and manual dexterity. ## Implementation hints Discrete graph traversal with stochastic edge resolution. Can be easily implemented as an adjacency matrix with difficulty weights. First milestone: construct the 8-state transition graph with deterministic transitions, then layer on skill parameters and cycle scaling.

Download as JSON

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

Download JSON
Cite this page

Ayatori (Japan). (n.d.). Games of Countries. https://gamesofcountries.com/games/jp-japan/ayatori

@misc{gamesofcountries-jp-japan-ayatori,
  title = {Ayatori (Japan)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/jp-japan/ayatori}
}