Skip to main content
Formal Specification

Tin Can Dodgeball — Formal Specification

Singapore · dexterity-throw · Confidence: high

← Back to game page

Implementation Suitability

Suitability: adaptation-requiredFidelity: probabilistic-proxyGap-fills: 3
# Bola Tin — AI-optimized specification ## General idea Bola Tin is an asymmetric multi-agent team challenge on a 2D discrete grid where a team of Builders attempts to assemble a pyramid of scattered blocks in the center of the field while evading a fielding team that passes a ball and fires targeted projectile attacks to eliminate the builders. ## Formal model - **Abstraction taken:** Continuous physical movement and projectile physics are discretized onto a 15×15 grid. The round progresses in discrete time ticks (sub-second steps). - **Player/agent count:** 2 teams of 3 agents each (Team B: Builders, Team F: Fielders). - **Turn model:** Simultaneous tick execution: all agents submit a directional movement, can-pickup/stack, or ball-pass/throw action per tick. ## Formal rules R1. Setup: A 15×15 grid. The pyramid base is at center (7,7). At tick 0, cans C=6 are scattered to random adjacent cells. Team B starts near (7,14); Team F starts distributed along edges. Ball starts at (7,7). R2. Simultaneous Movement & Actions per Tick: - Movement: Each agent can move 1 cell in cardinal directions (or stay). - Builder actions: `move`, `pickup_can`, `stack_can(at (7,7))`. Stacking requires an agent at (7,7) carrying a can. A stacked can increments `pyramid_progress` by 1. - Fielder actions: `move`, `pass_ball(target_teammate)`, `throw_ball(target_cell)`. R3. Combat / Dodge Resolution: - When a fielder with the ball executes `throw_ball(target_cell)`: - Trajectory is traced from fielder to target cell. - If a builder occupies the target cell (or intersecting path), a dodge check is evaluated: P(dodge)=clamp(0.1,0.8,builder_agility−thrower_accuracy). - If dodge fails, the builder is eliminated (*mati*) and removed from the grid. - Ball comes to rest at target cell (or nearest bound). R4. Ball Possession: A fielder moving into the ball's cell acquires possession. Builders cannot pick up the ball. R5. Win Conditions per Round: - Builder Victory: If `pyramid_progress` reaches 6, Team B wins the round immediately (+1 point). - Fielder Victory: If all 3 Builders are eliminated, Team F wins the round (0 points for B, roles swap). R6. Match Structure: The match lasts Rmax⁡=6 rounds. [gap-fill] A maximum of 100 ticks per round is enforced; if 100 ticks elapse without pyramid completion, Team F is awarded the round win. ## Edge cases - Tie at the end condition: If both teams have equal points after 6 rounds, a sudden-death speed-build challenge (time to stack 3 cans without defense) decides the tie [gap-fill]. - A player with no legal move: Not applicable; grid movement or wait actions are always available. - Repetition / cycles: Bounded by the 100-tick round cap [gap-fill]. - Simultaneous claims on the same resource or position: - If two fielders reach the ball simultaneously, the fielder with the lower agent index acquires it. - If a builder is stacking at (7,7) on the exact same tick a ball hits that cell, stacking evaluates first, but if the throw hits, the builder is eliminated after can placement. - Player elimination mid-round: Eliminated builders are immediately removed; the remaining builders continue play. - Resource exhaustion: Cans are invariant (6 cans total). - Termination guarantee: Every round terminates within at most 100 ticks, and the match terminates in 6 rounds. ## State representation - Persistent state: - `team_scores`: array of size 2 (integers ≥0). - `round_number`: integer 1..6. - Per-round state: - `tick_counter`: integer 0..100. - `pyramid_progress`: integer 0..6. - `builder_positions`: array of 3 coordinates (or null if eliminated). - `builder_carried_cans`: array of 3 booleans. - `fielder_positions`: array of 3 coordinates. - `ball_pos`: 2D coordinate (x,y). - `ball_carrier`: integer ∈{0,1,2,3} (0 = free ball). - Observable information: Fully observable to all agents. ## Action space - Builders: `move(direction)`, `pickup_can()`, `stack_can()`. - Fielders: `move(direction)`, `pass_ball(fielder_id)`, `throw_ball(target_pos)`. ## Playability notes - Single-player-vs-AI: Can be configured as 1 human builder + 2 AI builders vs. 3 AI fielders. - Group games (n>4): Naturally reduces to 2v2 or expands to 5v5 by adjusting grid dimensions and can counts. - Physical-skill games: Throwing accuracy and diving evasions are mapped to discrete stochastic dodge formulas. - What is lost: Physical pain/sting of being pelted with tennis balls, and vocal shouting coordination across the yard. ## Implementation hints Discrete grid multi-agent environment with simple line-of-sight raycasts for throws. Easily wrapped in an OpenAI Gym / PettingZoo environment.

Download as JSON

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

Download JSON
Cite this page

Tin Can Dodgeball (Singapore). (n.d.). Games of Countries. https://gamesofcountries.com/games/sg-singapore/bola-tin

@misc{gamesofcountries-sg-singapore-bola-tin,
  title = {Tin Can Dodgeball (Singapore)},
  year = {n.d.},
  publisher = {Games of Countries},
  url = {https://gamesofcountries.com/games/sg-singapore/bola-tin}
}