Skip to main content
Formal Specification

Four-Stick Board Race Game — Formal Specification

North Korea · board-race · Confidence: high

← Back to game page

Implementation Suitability

Suitability: engine-readyFidelity: directGap-fills: 1
# Yut nori — AI-optimized specification ## General idea A discrete turn-based circuit race game played on a 29-node graph using 4 two-sided casting lots. Players roll the sticks to earn movement values (1 to 5 stations), choose which token to advance around the circular perimeter or across inner diagonal shortcuts, and attempt to capture opponent pieces and stack friendly pieces to exit all tokens first. ## Formal model - **Abstraction taken:** None required for the core game logic; the casting lot outcome distribution is directly formalized based on physical stick probabilities. - **Player/agent count:** 2 opposing teams (Team 0, Team 1), each controlling 4 tokens (or k tokens in mass configurations). - **Turn model:** Sequential alternating turns with bonus rolls for rolling *Yut* (4), *Mo* (5), or capturing an opponent token. ## Formal rules R1. Setup: Board is a directed graph with 29 nodes: 20 perimeter stations `N0..N19`, 8 diagonal stations `D1..D8`, and 1 center hub `C`. Each team starts with 4 tokens in reserve (`Off-Board`). R2. On a turn, the active team casts 4 binary lots. Outcome distribution: - 1 flat up (*Do*): value = 1. - 2 flat up (*Gae*): value = 2. - 3 flat up (*Geol*): value = 3. - 4 flat up (*Yut*): value = 4, grant 1 bonus cast. - 0 flat up (*Mo*): value = 5, grant 1 bonus cast. R3. All earned casts are accumulated into a multiset of move values before (or sequentially during) token movements. R4. For each available move value, the player selects one active token on board or in reserve to advance along legal directed edges. R5. Routing logic: - Normal perimeter path: `N0 -> N1 -> ... -> N19 -> Goal`. - Corner shortcut 1: Landing exactly on `N5` redirects future movement through `D1 -> D2 -> C`. - Corner shortcut 2: Landing exactly on `N10` redirects through `D3 -> D4 -> C`. - Center hub routing: Landing on `C` from `N5` routes to `D5 -> D6 -> N15 -> Goal`; from `N10` routes to `D7 -> D8 -> N0 -> Goal`. R6. **Capture (*Japgi*):** If a token finishes its move on an opponent-occupied node, the opponent token(s) are returned to reserve (`Off-Board`), and the active team immediately receives 1 extra roll. R7. **Stacking (*Eopgi*):** If a token lands on a friendly token, they merge into a single stack and move together as one piece on all subsequent moves. R8. Goal condition: A token/stack that moves past the final node exits the board (`Finished`). R9. The first team to have all 4 tokens in `Finished` state immediately wins the game. R10. [gap-fill] Loop-breaking move limit: If total turns reach 300 without a winner, the game is declared a draw, or decided by the team with the most exited/advanced tokens. ## Edge cases - **Tie at the end condition:** Not applicable; turns are strictly sequential, so one team achieves all tokens exited first. - **A player with no legal move:** If a team has available move values but all valid moves are blocked by specific variant rules (rare), unused move values are discarded. - **Repetition / cycles:** Tokens can loop around the board repeatedly if they miss exit corners; bounded by R10's 300-turn cap. - **Simultaneous claims on the same resource or position:** Sequential turns prevent simultaneous claims; landing on an occupied node triggers deterministic capture (R6) or stacking (R7). - **Player elimination mid-round:** Not applicable (team-level token race). - **Resource exhaustion:** Tokens are never permanently destroyed; captured tokens re-enter from reserve. - **Termination guarantee:** Guaranteed by finite board length, absorbing Goal state, and R10 turn cap. ## State representation - **Persistent state:** `token_positions[team_id][token_id]` in `{Off-Board, Node_ID, Finished}`, `stack_clusters`. - **Per-turn state:** `active_team`, `pending_roll_values` (list of integers `[1..5]`), `bonus_roll_count`. - **Observable information:** Fully observable perfect-information game. ## Action space - `cast_lots()`: Sample random lot outcome. - `move_token(token_id, value)`: Apply specified value to chosen token along valid board path. ## Playability notes - **Single-player-vs-AI:** Highly engine-ready; easily modeled with standard Expectiminimax or Monte Carlo Tree Search (MCTS). - **Group games (n>4):** Reduces cleanly to 2 competing sides without altering the underlying board state mechanics. ## Implementation hints Discrete state space. The 29-node board graph with branching paths is simple to represent as an adjacency list. Branching factor is at most 4 token choices per roll value.

Download as JSON

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

Download JSON
Cite this page

Four-Stick Board Race Game (North Korea). (n.d.). Games of Countries. https://gamesofcountries.com/games/kp-north-korea/yut-nori

@misc{gamesofcountries-kp-north-korea-yut-nori,
  title = {Four-Stick Board Race Game (North Korea)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/kp-north-korea/yut-nori}
}