Formal Specification
Clay Pot Breaking (also known as Veay Ka-Am Boran) — Formal Specification
Cambodia · dexterity-throw · Confidence: low
← Back to game pageImplementation Suitability
Suitability: engine-readyFidelity: probabilistic-proxyGap-fills: 2
# Veay Ka-Am — AI-optimized specification
## General idea
A single-agent blindfolded navigation and target-striking simulation modeling clay pot breaking. The contestant's spatial orientation is scrambled by three initial spins, requiring the agent to filter noisy crowd directional audio cues (a mixture of authentic guidance and deceptive shouts) to take up to three walking strides and execute a single decisive stick swing at a suspended earthenware jar.
## Formal model
- **Abstraction taken:** Blindfolded physical balance, auditory triangulation, and stick-swing kinematics are abstracted into a 2D spatial belief state with noisy directional observations and a Euclidean hit-box collision resolver.
- **Player/agent count:** 1 active contestant agent per heat, evaluated in sequence across tournament players.
- **Turn model:** Sequential single-agent multi-step heat (at most 3 movement strides followed by 1 swing).
## Formal rules
R1. Arena Setup: Target terracotta pot is suspended at fixed 2D coordinate meters with effective strike radius meters.
R2. Initialization & Disorientation:
- Contestant starts at origin .
- Heading angle is randomized uniformly over to simulate the three disorientation spins (*bangvel bei chhum*).
- Contestant internal state maintains a belief distribution over position and heading: .
R3. Movement & Observation Strides:
- The contestant takes up to walking strides.
- On each stride , the agent executes `step_stride(step_length, turn_angle)` where meters and .
- [gap-fill] Noisy Auditory Observation: After each stride, the agent receives an observed angle cue sampled from a Gaussian mixture:
- With (helpful crowd shouting): with rad.
- With (deceptive crowd shouting): with rad.
R4. Strike Execution:
- After stride 3 (or voluntarily earlier), the contestant commits to a single swing action `swing_stick(aim_angle, reach_distance)` where meters.
- [gap-fill] Collision Evaluation:
- The stick impact point is calculated as .
- If : The pot shatters Contestant scores 10 points (and claims prize reward).
- If : Swing misses empty air Contestant scores 0 points.
R5. Tournament Win Condition: After all contestants complete their heats, the contestant with the highest score (or lowest miss distance as tie-breaker) wins the tournament.
## Edge cases
- **Tie at the end condition:** If multiple contestants shatter their pots, the tie is broken by fewest strides taken; if still tied, a sudden-death heat with a smaller pot (m) is staged.
- **A player with no legal move:** Not applicable — stride and swing actions are always available within parameter bounds.
- **Repetition / cycles (and the loop-breaking rule):** Bounded by strict stride counter and mandatory terminal swing.
- **Simultaneous claims on the same resource or position:** Not applicable — contestants perform in sequential isolated heats.
- **Player elimination mid-round; turn order repair:** Not applicable — each contestant receives exactly one complete heat.
- **Resource exhaustion:** Earthenware pots are replaced fresh for each contestant heat.
- **Termination guarantee: why the game provably ends:** Guaranteed because every heat executes at most 3 strides and exactly 1 swing, terminating in finite steps.
## State representation
- `true_pos`: float coordinate .
- `true_heading`: float angle .
- `stride_count`: integer in .
- `has_swung`: boolean.
- `score`: integer .
- Observable information: Information set is strictly imperfect — agent does NOT observe `true_pos` or `true_heading`; observes only noisy auditory signals and step odometry.
## Action space
- `step_stride(step_length, turn_angle)`: where , .
- `swing_stick(aim_angle, reach_distance)`: where , .
## Playability notes
- **Single-player-vs-AI:** Excellent benchmark for Partially Observable Markov Decision Processes (POMDP) and Bayesian filtering algorithms (Kalman filter / particle filter).
- **Group games (n>4):** Evaluated as a tournament leaderboard across individual agent heats.
- **What is lost:** The festive visual humor of blindfolded stumbling, sensory disorientation, and communal laughter are modeled mathematically as belief-state uncertainty.
## Implementation hints
- Implement the agent's internal state as a particle filter updating belief distribution over upon receiving noisy crowd cues .
- Entire simulation engine is under 70 lines of code.
Download as JSON
Download the full game corpus entry for this game as structured JSON.
Download JSONCite this page
Clay Pot Breaking (also known as Veay Ka-Am Boran) (Cambodia). (n.d.). Games of Countries. https://gamesofcountries.com/games/kh-cambodia/veay-ka-am
@misc{gamesofcountries-kh-cambodia-veay-ka-am,
title = {Clay Pot Breaking (also known as Veay Ka-Am Boran) (Cambodia)},
year = {n.d.},
publisher = {Games of Countries},
url = {https://gamesofcountries.com/games/kh-cambodia/veay-ka-am}
}