Formal Specification
Khan's Encampment — Formal Specification
Kyrgyzstan · dexterity-throw · Confidence: low
← Back to game pageImplementation Suitability
Suitability: engine-readyFidelity: probabilistic-proxyGap-fills: 1
# Ordo — AI-optimized specification
## General idea
A turn-based team tactical projectile game played in a large 14-meter circle simulating siege warfare. Players use heavy knucklebone discs (*tompoi*) to dislodge sheep knucklebones (*chükö*) and a central Khan from a fortified perimeter. It differs from standard lawn bowls or curling by requiring pieces to be completely ejected beyond a boundary line, having strict restrictions on targeting the high-value Khan until defenses are depleted, and terminating turns immediately upon fouls or missed dislodgements.
## Formal model
- **Abstraction taken:** The physical finger flick, sliding spin, and lofted ballistic trajectory of the *tompoi* on packed clay are abstracted into a 2D physics collision vector (aim angle, shot power, shot type) combined with a stochastic Gaussian deviation based on agent precision skill.
- **Player/agent count:** 2 teams of 5 agents (10 total), alternating turns in sequential rotation.
- **Turn model:** Sequential shooting turns. A player retains their turn as long as each shot legally expels at least one piece without fouling; a miss or foul passes the turn to the opposing team.
## Formal rules
R1. Setup: Circle of radius R = 7.0m centered at (0,0). Central pit at (0,0) holds the Khan piece. Exactly 50 sheep knucklebones (*chükö*) initialized in concentric cluster around the Khan within radius r = 0.6m.
R2. Shot Selection: Active shooter chooses a shot technique:
- `power_strike(angle, power)`: High velocity perimeter strike aimed at center pack.
- `pinning_shot(target_id, angle, power)`: Precision slide shot aimed at an individual perimeter piece.
- `finger_snap(target_id)`: Short-range snap from inside circle.
R3. Physics & Expulsion:
- Trajectories and elastic collisions are simulated.
- Any *chükö* that crosses distance `d > 7.0m` from center is removed from play and credited to the shooting team (+1 point).
- The *tompoi* must not cross the boundary if shot from inside, and must not remain dead inside the circle without expelling a piece.
R4. Khan Targeting: The Khan may only be targeted if the total remaining *chükö* count on the board is <= 10 (or central garrison is cleared). Striking the Khan out of bounds awards +10 points (or instant victory).
R5. Turn Continuation & Stoppage (*Жыдымай*):
- If a shot expels >= 1 piece legally, the shooter may shoot again.
- If a shot fails to expel a piece, or incurs a foul (boundary violation, illegal Khan touch), the turn ends immediately and passes to the next player in rotation from the opposing team.
R6. Win Condition: The game ends when all pieces are cleared, or after 10 full rotation innings. The team with the highest point total wins.
## Edge cases
- Tie at the end condition: [gap-fill] 3-shot sudden-death shootout between team captains from the perimeter.
- A player with no legal move: Not applicable; shooting into the circle is always legal.
- Repetition / cycles: Bounded because pieces are permanently removed on expulsion; maximum possible shots is strictly finite.
- Simultaneous claims on the same resource: Not applicable; turns are strictly sequential.
- Player elimination mid-round: If a player fouls out, their position in rotation is skipped.
- Resource exhaustion: Monotonically decreasing piece count guarantees end state when pieces reach 0.
- Termination guarantee: Upper bound of 10 innings combined with finite piece pool guarantees termination.
## State representation
- Persistent state: Piece positions (2D coordinates for Khan and up to 50 *chükö*), piece alive status (boolean array), score per team, current inning (1..10), active team, active player index.
- Ephemeral state: Last shot outcome vector, list of pieces expelled this turn.
- Observability: Complete geometric information, fully observed by all agents.
## Action space
- `shoot(technique_id, angle_rad, power_norm, target_piece_id)`: Discretizable into 36 angle directions and 5 power levels per technique.
## Playability notes
- Single-player-vs-AI: Highly playable; human takes shots with mouse aim/power meter while AI calculates tactical trajectories using 2D physics forward-projection.
- Group games: Easily compressed to 1v1 where each player controls an entire 5-man team.
- Physical skill: Parameterized via Gaussian aim dispersion `sigma = f(agent_skill)`.
- What is lost: The fine tactile feel of ground unevenness, finger-callous release, and the heavy bone weight of the *tompoi*.
## Implementation hints
Can be built as a clean 2D rigid-body disc physics engine (e.g. standard circle-circle elastic collision). Branching factor is continuous but discretizes to ~100 candidate target vectors. First milestone: implement 2D circle boundary with 10 knucklebones and single-shot expulsion scoring before scaling to full 50-bone garrison.
Download as JSON
Download the full game corpus entry for this game as structured JSON.
Download JSONCite this page
Khan's Encampment (Kyrgyzstan). (n.d.). Games of Countries. https://gamesofcountries.com/games/kg-kyrgyzstan/ordo
@misc{gamesofcountries-kg-kyrgyzstan-ordo,
title = {Khan's Encampment (Kyrgyzstan)},
year = {n.d.},
publisher = {Games of Countries},
url = {https://gamesofcountries.com/games/kg-kyrgyzstan/ordo}
}