Skip to main content
Formal Specification

Throwing the Stone from the Shoulder — Formal Specification

Montenegro · strength-endurance · Confidence: high

← Back to game page

Implementation Suitability

Suitability: adaptation-requiredFidelity: abstractedGap-fills: 2
# Bacanje kamena s ramena — AI-optimized specification ## General idea A traditional Highland stone-putting contest where agents compete sequentially across fixed rounds to record the longest valid forward distance. The strategic layer centers on balancing release angle, approach velocity, and baseline proximity to maximize distance while avoiding foul oversteps. The continuous biomechanics of shoulder thrust and foot-fault risk are abstracted into a parameterized stochastic distance sampler. ## Formal model - **Abstraction taken:** The continuous physical throw is abstracted into a discrete decision over an intended effort/risk level, resolving into a sampled distance `D ~ Normal(mu(skill, effort), sigma)` and a boolean foul probability `P(fault) = f(effort, skill)`. - **Player/agent count:** 2 to n agents (standard 8-agent tournament bracket). - **Turn model:** Strictly sequential across 3 fixed preliminary rounds. ## Formal rules R1. Setup: A tournament consists of `N` registered agents and a fixed schedule of `R = 3` throwing rounds. R2. Turn order: Agents throw in fixed ascending index order `1..N` in each round. R3. Legal action: On an agent's turn, the active agent selects an action `throw(effort_level)` where `effort_level in [0.0, 1.0]`. R4. Resolution: - [gap-fill] A foul check is evaluated: with probability `P(fault) = baseline_risk + effort_level * foul_sensitivity`, the attempt is ruled a fault (`distance = 0.0`). - If no foul occurs, the distance is sampled from `D = max(0.0, base_distance * (0.7 + 0.3 * effort_level) + noise)`. R5. State update: The agent's best distance `best_score[agent]` is updated to `max(best_score[agent], D)`. R6. End condition: After all `N` agents complete `R` rounds, the agent with the highest `best_score` wins. R7. [gap-fill] Tie-breaker: If two or more top agents share the exact same maximum distance, they participate in a single sudden-death jump-off throw round until a unique maximum is established. ## Edge cases - Tie at the end condition: Resolved via R7 (sudden-death round among tied top agents). - A player with no legal move: Not applicable; `throw(effort)` is always legal on an active turn. - Repetition / cycles: Bounded strictly by the fixed 3-round structure; jump-off ties break with continuous stochastic outcomes having probability 0 of identical real values. - Simultaneous claims on the same resource or position: Not applicable; throws are executed in strict sequential order. - Player elimination mid-round; turn order repair: If an agent disconnects or forfeits mid-tournament, their remaining throws score `0.0`, and turn rotation continues with remaining active agents. - Resource exhaustion: Not applicable; stone is retrieved after each throw, and rounds are fixed. - Termination guarantee: Provably terminates in exactly `N * 3` turns (plus rare single-throw tie-breakers). ## State representation - Persistent state: `best_score: map`, `current_round: int (1..3)`, `current_thrower_idx: int (0..N-1)`. - Per-turn ephemeral state: Selected `effort_level`, sampled `is_foul: bool`, sampled `distance: float`. - Observable information: Fully observable public state (all throws, fouls, and distances are publicly broadcast). ## Action space - `throw(effort_level: float)`: Parameter `effort_level in [0.0, 1.0]`, representing conservative safe throws vs. aggressive maximum-distance throws with elevated fault risk. ## Playability notes - Single-player-vs-AI: Natural; 1 human vs. 1..n AI agents with differing skill profiles. - Group games (n>4): Fully scalable; leaderboard tracks best distance per agent without coalition interference. - Physical-skill games: Skill modeled via two parameters per agent: `power` (shifts mean distance upward) and `technique` (reduces foul probability and outcome variance). - What is lost: The intricate muscle kinesthetics, explosive footwork, and physical fatigue of hoisting a 12 kg stone are abstracted into scalar probability distributions. ## Implementation hints Extremely lightweight state machine. Branching factor is 1 continuous parameter per turn (or a discretized choice of 3–5 risk tiers: Safe, Balanced, Maximum). First milestone: implement sequential 3-round leaderboard with stochastic resolver and verify tie-break termination.

Download as JSON

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

Download JSON
Cite this page

Throwing the Stone from the Shoulder (Montenegro). (n.d.). Games of Countries. https://gamesofcountries.com/games/me-montenegro/bacanje-kamena-s-ramena

@misc{gamesofcountries-me-montenegro-bacanje-kamena-s-ramena,
  title = {Throwing the Stone from the Shoulder (Montenegro)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/me-montenegro/bacanje-kamena-s-ramena}
}