Skip to main content
Formal Specification

Wolf and marmots — Formal Specification

Mongolia · chase-tag · Confidence: low

← Back to game page

Implementation Suitability

Suitability: engine-readyFidelity: discretizedGap-fills: 1
# Chono tarvagatsakh — AI-optimized specification ## General idea An asymmetric multi-agent predator-prey tag game on a 2D grid. One Wolf agent starts in a central den, while several Marmot agents start in perimeter burrow cells. Marmots must forage in open pasture cells to accumulate grazing points, but must sprint back to burrow sanctuaries before the faster Wolf agent ambushes and tags them. ## Formal model - **Abstraction taken:** The playing field is discretized into a 2D grid `[-10, 10] x [-10, 10]`. Central cell `(0, 0)` is the Wolf Den. Several designated peripheral cells `B = {b_1, ..., b_k}` are Marmot Burrows with a fixed maximum capacity of 2 marmots. - **Player/agent count:** 1 Wolf agent vs. `M` Marmot agents (typically 6 to 10 marmots). - **Turn model:** Synchronized simultaneous ticks `t = 1, 2, ...`. ## Formal rules R1. Setup: Wolf starts at `(0, 0)` with `speed_wolf = 2` cells/tick. Marmots start in burrow cells `B` with `speed_marmot = 1` cell/tick. R2. Phase 1 — Foraging & Dialogue: - For the first `T_dialogue = 5` ticks, the Wolf is dormant in its den ("sleeping"). - Marmots move outward into open pasture cells `(x, y) not in B`, scoring +1 foraging point for every tick spent outside a burrow. R3. Phase 2 — Ambush & Chase: - At tick `T_dialogue + 1`, the Wolf awakens and pursues active Marmots. - On each tick, all agents simultaneously submit a directional move vector `(dx, dy)`. R4. Burrow Sanctuary & Overcrowding: - A Marmot occupying a burrow cell `b in B` is completely immune to tagging. - [gap-fill] Capacity rule: A burrow cell can shelter at most 2 Marmots simultaneously. If a 3rd Marmot enters, the longest-resident Marmot in that cell is displaced 1 cell outward into the open. R5. Tagging: - If the Wolf moves into the same cell as an unshielded Marmot in the open pasture, that Marmot is tagged. - In elimination format, the tagged Marmot is captured into the Wolf's den and removed from active play. R6. Win condition: - Wolf wins a round if it tags all Marmots before `T_max = 50` ticks. - Marmots win if any Marmots survive until `T_max`, with individual Marmots ranked by accumulated foraging points. ## Edge cases - Tie at the end condition: Resolved by individual Marmot foraging point totals. - A player with no legal move: Not applicable — 8-directional grid movement is always available. - Repetition / cycles: Bounded strictly by `T_max = 50` ticks per round. - Simultaneous claims on the same resource: If Wolf and Marmot move onto the same cell on the same tick, tag resolves immediately. - Player elimination mid-round: Tagged marmots are removed from the active movement queue; remaining marmots continue playing. - Resource exhaustion: Fixed duration round (50 ticks) terminates cleanly. - Termination guarantee: Provably terminates at tick `T_max = 50`. ## State representation - Persistent state: `pos_wolf` `(x, y)`, `marmot_states` (array of structs `{id, pos, status in {ACTIVE, SAFE_IN_BURROW, CAPTURED}, foraging_points}`), `burrow_occupancy` (map of burrow_id -> list of marmot IDs), `tick_counter` (0..50). - Per-tick ephemeral state: Submitted move vectors, tag collision detections, displaced marmot events. - Observable information: Fully observable — all agent positions, speeds, burrow locations, and score tallies are public. ## Action space - `move_wolf(dx, dy)`: Step up to 2 cells (Chebyshev distance `<= 2`). - `move_marmot(dx, dy)`: Step up to 1 cell (Chebyshev distance `<= 1`). ## Playability notes - Single-player-vs-AI: Natural — 1 human player controlling either the Wolf (optimizing pursuit trajectories) or a lead Marmot (optimizing greedy foraging vs. burrow proximity). - Group games (n>4): Multi-agent pursuit-evasion benchmark environment. - Physical-skill games: Deterministic pursuit-evasion grid problem. - What is lost in the adaptation: The theatrical singing of the rhyming dialogue (*"Chono guai, chono guai, tsog ögөөch!"*) and children's laughter. ## Implementation hints Classic pursuit-evasion grid game (similar to Pac-Man or Predator-Prey multi-agent RL). Milestone 1: implement grid, burrow sanctuary check, and Manhattan distance pursuit heuristic for Wolf; Milestone 2: implement multi-marmot cooperative evasion policy.

Download as JSON

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

Download JSON
Cite this page

Wolf and marmots (Mongolia). (n.d.). Games of Countries. https://gamesofcountries.com/games/mn-mongolia/chono-tarvagatsakh

@misc{gamesofcountries-mn-mongolia-chono-tarvagatsakh,
  title = {Wolf and marmots (Mongolia)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/mn-mongolia/chono-tarvagatsakh}
}