Formal Specification
The Bars (Prisoners' Base) — Formal Specification
Jersey · chase-tag · Confidence: low
← Back to game pageImplementation Suitability
Suitability: engine-readyFidelity: discretizedGap-fills: 2
# Les Barres — AI-optimized specification
## General idea
A dynamic multi-agent tag and territory game where legal capture authority depends on temporal priority ("freshness"). Two teams operate from home bases at opposite ends of a 2D grid. Stepping off the home line records an entry timestamp; an agent can tag any opposing agent whose active timestamp is strictly older, creating an emergent cat-and-mouse dynamic of baiting, refreshing, and rescuing captured teammates from prison.
## Formal model
- **Abstraction taken:** The continuous field is discretized into a 2D grid ( cells). Time progresses in discrete ticks . Continuous movement is replaced by 1-cell orthogonal or diagonal moves per tick.
- **Player/agent count:** 2 teams (Team and Team ), with agents per team (standard) or agent per team (minimal tactical duel).
- **Turn model:** Simultaneous-move execution per tick: all agents submit movement/tag actions, which are validated and resolved in order of freshness.
## Formal rules
R1. Grid Topology: The board is an grid ().
- Team Home Base: . Team Prison: .
- Team Home Base: . Team Prison: .
R2. Agent State Vector: Each agent has state where:
- .
- is the discrete tick index when the agent last transitioned from to .
R3. Movement: At each tick, active agents in or may move to an adjacent cell (Chebyshev distance ).
- Moving into one's own home base resets status to and clears .
- Exiting one's home base into sets status to and updates .
R4. Tagging and Freshness:
- If agent and agent occupy the same cell or adjacent cells (distance ):
- If , agent captures .
- transitions to at Team 's prison location .
- If [gap-fill: simultaneous exit results in mutual bounce with no capture].
R5. Prison and Rescue:
- Multiple prisoners at a prison form a chain: Prisoner at distance from the prison anchor.
- If an active teammate in reaches a cell adjacent to any member of the prisoner chain, all prisoners in that chain and the rescuer transition to .
- Agents in move directly toward their home base and cannot tag or be tagged.
R6. Win Conditions:
- A team wins immediately if all opposing agents are concurrently in state.
- Alternatively, an active agent who steps onto the opponent's home base line without being tagged scores an immediate win for their team.
R7. [gap-fill] Max-Tick Termination: If neither win condition is achieved by ticks, the team with fewer active prisoners wins; if equal, the match is a draw.
## Edge cases
- Tie at the end condition: Resolved via R7 (fewer prisoners, then draw).
- A player with no legal move: Agents in prison have no movement actions until rescued; agents at base can always remain at base or exit.
- Repetition / cycles (and the loop-breaking rule): Bounded by the hard tick horizon .
- Simultaneous claims on the same resource or position: Simultaneous entry to a cell by two opposing players is resolved by freshness check (); if timestamps are identical, both agents bounce back to their previous tick coordinates.
- Player elimination mid-round; turn order repair: Captured agents enter prison state and are skipped in active field collision checks until liberated.
- Resource exhaustion (pieces, tokens, space, time): Grid cells and agent counts are constant; time is bounded by .
- Termination guarantee: Guaranteed by the hard cap combined with immediate terminal states in R6.
## State representation
- Persistent state:
- `tick`: Integer counter .
- `agents`: Array of 8 agent objects .
- Ephemeral state (per tick):
- `declared_moves`: Map of agent ID .
- Observable information: Fully observable game; all agent coordinates, team affiliations, statuses, and entry timestamps are known to all agents.
## Action space
- `move(dx, dy)`: . Valid for and .
- `stay_at_base()`: Valid for .
- `step_out()`: Transitions from to , updating .
## Playability notes
- Single-player-vs-AI: Highly playable; human commands one agent (or a squad) against an AI opponent. The AI maintains a priority queue of enemy freshness timestamps and calculates safety corridors to enemy prison/base.
- Group games (n>4): The multi-agent form cleanly abstracts full squad dynamics without rules modification.
- Physical-skill games: Physical sprinting speed is abstracted into uniform grid movement speed; the game becomes a pure game-theoretic spatial coordination and timing problem.
- What is lost in the adaptation: Acceleration bursts, feints, physical reach, and simultaneous multi-body tag collisions.
## Implementation hints
- Freshness tracking is trivial: a single integer `last_exit_tick` per agent.
- State evaluation function for AI agents can compute: .
- Recommended first milestone: Implement a 1v1 duel on a grid to verify freshness capture logic and base touching before expanding to with prison rescue chains.
Download as JSON
Download the full game corpus entry for this game as structured JSON.
Download JSONCite this page
The Bars (Prisoners' Base) (Jersey). (n.d.). Games of Countries. https://gamesofcountries.com/games/je-jersey/les-barres
@misc{gamesofcountries-je-jersey-les-barres,
title = {The Bars (Prisoners' Base) (Jersey)},
year = {n.d.},
publisher = {Games of Countries},
url = {https://gamesofcountries.com/games/je-jersey/les-barres}
}