Formal Specification
Monkey on the Peg / Slipper Guard — Formal Specification
Pakistan · chase-tag · Confidence: low
← Back to game pageImplementation Suitability
Suitability: engine-readyFidelity: discretizedGap-fills: 3
# Bandar Killa — AI-optimized specification
## General idea
An asymmetric 1-vs-many perimeter defense and resource retrieval game. A central Guard agent tethered to origin `(0, 0)` by a fixed radius rope must defend a central cache of `2*(N-1)` resource tokens (shoes) from `N-1` surrounding Retriever agents. If the Guard tags any Retriever, roles invert immediately. If all tokens are stolen, a secondary sprint gauntlet phase initiates where the Guard attempts to reach a safe sanctuary line.
## Formal model
- **Abstraction taken:** The 2D continuous playground is discretized into coordinates `(x, y)`. The tether post is at `(0, 0)` with maximum guard radius `R_rope = 1.5` meters. The safe sanctuary line is located at `x = 15.0`.
- **Player/agent count:** `N` agents total: 1 Guard (Monkey `M`), `N-1` Retrievers `R_1..R_{N-1}` (standard `N = 5`, total 8 shoes).
- **Turn model:** Discrete real-time simulation ticks `Delta_t = 0.2` seconds.
## Formal rules
R1. Setup: Peg at `(0, 0)`. Initial shoe pile count `S = 2 * (N - 1)`. Guard starts at `(1.0, 0.0)` with constraint `x_M^2 + y_M^2 <= R_rope^2`. Retrievers start at distance `r = 4.0` meters from origin.
R2. Phase 1 (Guard & Steal): On each tick:
- Guard updates position: `||(dx_M, dy_M)|| <= v_guard` subject to `||(x_M + dx_M, y_M + dy_M)|| <= R_rope`.
- Retrievers update positions: `||(dx_i, dy_i)|| <= v_retriever`.
R3. Tag check: If `dist(M, R_i) <= r_tag` and `||pos_M|| <= R_rope`:
- Retriever `R_i` is tagged.
- Role inversion: `R_i` becomes the new Guard. Shoes reset to `S = 2 * (N - 1)`. Phase 1 restarts.
R4. Shoe retrieval: If an untagged retriever `R_i` is within `dist(R_i, (0, 0)) <= r_steal` and `S > 0`:
- Retriever executes `steal_shoe()` -> `S` decrements by 1 (`S -= 1`).
R5. Phase 2 (Escape Gauntlet): As soon as `S == 0`:
- Tether constraint is removed for Guard.
- Guard's objective: Reach `x >= 15.0` (Sanctuary line).
- Retrievers pursue Guard; if any retriever reaches `dist(R_i, M) <= r_throw`, retriever executes `throw_slipper(M)`, inflicting penalty points on Guard.
R6. Match end: When Guard crosses `x >= 15.0`, round terminates.
R7. [gap-fill] Score calculation: Guard earns `score = 100 - (penalty_hits * 10)`. Retrievers split stolen shoe bounty.
R8. [gap-fill] Turn rotation: Each player takes 1 turn as initial Guard. After `N` rounds, the agent with the highest individual score wins.
## Edge cases
- **Tie at the end condition:** Handled by point accumulation across `N` symmetric rounds (R8).
- **A player with no legal move:** Spatial displacement vectors always admit `(0, 0)` (stationary).
- **Repetition / cycles:** Guarded phase is bounded because retrievers coordinate multi-angle simultaneous rushes; if no progress after 300 ticks, [gap-fill] random shoe decay occurs.
- **Simultaneous claims on the same resource or position:** Multiple retrievers at `(0, 0)` steal 1 shoe each per tick up to remaining pile count `S`.
- **Player elimination mid-round:** Not applicable; players invert roles or transition to Phase 2.
- **Resource exhaustion:** Shoe pile `S` reaches 0, which triggers the deterministic state transition to Phase 2.
- **Termination guarantee:** Guaranteed by finite shoe cache `S` transitioning to a unidirectional sprint to `x = 15.0`.
## State representation
- **Persistent state:** `scores[1..N]`, `round_index in {1..N}`, `active_guard_id in {1..N}`.
- **Per-tick ephemeral state:** `phase in {Phase_1_Guard, Phase_2_Gauntlet}`, `shoes_remaining S in {0..2(N-1)}`, positions `(x, y)` for all `N` agents, `hit_counter`.
- **Observable information:** Fully observable 2D state to all agents.
## Action space
- `move(dx, dy)`: Displacement vector.
- `steal_shoe()`: Retriever action at origin.
- `tag_retriever(id)`: Guard offensive action.
- `throw_slipper(target_id)`: Phase 2 retriever action.
## Playability notes
- **Single-player-vs-AI:** Natural; human player can control the tethered Guard trying to anticipate AI multi-agent flanking attacks, or play as a daring Retriever.
- **Group games (n>4):** Native asymmetric multiplayer design (1 vs 4+).
- **Physical-skill games:** Speed, reach, and rope-tension mechanics are modeled via radius constraints and collision radii.
- **What is lost:** The kinetic tension of dodging physical shoe throws and the physical feel of pulling a shoe from the dust while laughing.
## Implementation hints
Low complexity. The circular tether constraint is mathematically trivial (`clamp(pos, max_radius)`). Implement Phase 1 first with 1 Guard and 4 AI Retrievers before adding Phase 2 sprint logic.
Download as JSON
Download the full game corpus entry for this game as structured JSON.
Download JSONCite this page
Monkey on the Peg / Slipper Guard (Pakistan). (n.d.). Games of Countries. https://gamesofcountries.com/games/pk-pakistan/bandar-killa
@misc{gamesofcountries-pk-pakistan-bandar-killa,
title = {Monkey on the Peg / Slipper Guard (Pakistan)},
year = {n.d.},
publisher = {Games of Countries},
url = {https://gamesofcountries.com/games/pk-pakistan/bandar-killa}
}