Formal Specification
Kanamachi (Blind Bee) — Formal Specification
Bangladesh · chase-tag · Confidence: high
← Back to game pageImplementation Suitability
Suitability: engine-readyFidelity: discretizedGap-fills: 3
# Kanamachi — AI-optimized specification
## General idea
A traditional Bengali circle-tagging and sensory localization game. One player is blindfolded as the "Blind Bee" (*Kana*), standing at the center of a circular arena, while other players ("bees") circle around, teasing the catcher with rhythmic nursery rhymes (*"Kanamachi bho bho, jake pabi take chho!"*) and light taps. The blindfolded catcher must localize sounds and footfalls to lunge and grab a teasing player, transferring the blindfold role upon capture.
## Formal model
- **Abstraction taken:** The 2D circular courtyard (radius ) is modeled with continuous polar coordinates . The catcher's sensory impairment is modeled as a noisy auditory vector receiving directional angle and distance signals from teasing players. Physical grabbing and tactile identification are resolved via stochastic checks.
- **Player/agent count:** 5 to 8 players (1 Catcher and Bees ).
- **Turn model:** Real-time ticked / simultaneous turns per round (1 tick = 0.5 sec). A round concludes when the catcher successfully captures and identifies a bee. Match consists of 10 rounds.
## Formal rules
R1. Setup:
- Circular arena of radius centered at .
- Player starts blindfolded at .
- Circling Bees start distributed on the circle perimeter ().
R2. Circling & Teasing:
- On each tick, Bees submit movement vectors and may declare a `TEASE_TAP` action if within of .
- Each `TEASE_TAP` generates an auditory signal:
R3. Catcher Movement & Grab:
- rotates heading based on auditory signals and moves at blind walk speed ().
- may execute a `LUNGE_GRAB` action.
- [gap-fill] Grab Success Check: If any bee is within of and within the catcher's arm-sweep cone ( of heading):
R4. Tactile Identification Phase (*নাম অনুমান*):
- When a bee is grabbed, the round pauses.
- [gap-fill] touches the captive's head/shoulders and submits a guess .
- If correct: becomes the new for the next round; the previous catcher scores point and joins the circling bees.
- If incorrect: is released, receives a 5-second penalty spin, and the current round continues.
R5. Boundary Constraint: If a bee steps outside the arena radius, they are automatically penalized and become the new immediately.
R6. [gap-fill] Round Time Limit: If a round reaches 60 ticks (30 seconds) without a capture, the bee who executed the fewest `TEASE_TAP` actions is penalized to become the new .
R7. Win Condition: After 10 completed rounds, the player with the highest score (fewest rounds spent as / most successful identifications) wins the match.
## Edge cases
- Tie at the end condition: If players are tied in score after 10 rounds, the player who spent the least total ticks as wins.
- A player with no legal move: All players submit continuous movement vectors within polar bounds.
- Repetition / cycles: Bounded strictly by the 60-tick per-round cap (R6) and 10-round match horizon.
- Simultaneous claims on the same resource or position: If multiple bees are within the grab cone during a lunge, the closest bee is targeted.
- Player elimination mid-round; turn order repair: No elimination; roles rotate continuously among all active participants.
- Resource exhaustion (pieces, tokens, space, time): Round timer expiration triggers automatic role rotation (R6).
- Termination guarantee: Round time limits (60 ticks) guarantee that all 10 rounds complete in finite time.
## State representation
- Persistent state:
- `round_number`: Integer .
- `player_scores`: Map of player IDs to integer points.
- `time_as_catcher`: Map of player IDs to integer ticks spent blindfolded.
- `active_kana_id`: Current catcher player ID.
- Per-turn ephemeral state:
- `tick_counter`: Integer .
- `player_coords`: Map of player IDs to positions.
- `tease_counts`: Map of bee IDs to tease actions executed this round.
- `auditory_cues`: List of recent sound signals .
- Observable information: Bees have complete visual observability of the arena; has zero visual input and observes only noisy auditory sound vectors.
## Action space
- `move_bee(dr, dtheta)`: Reposition bee within arena boundaries.
- `tease_tap()`: Bee lightly taps catcher from close range.
- `orient_and_walk(dtheta, speed)`: Catcher rotates and steps forward.
- `lunge_grab()`: Catcher attempts to seize an adjacent bee.
- `identify_captive(suspect_id)`: Catcher declares name guess for grabbed captive.
## Playability notes
- Single-player-vs-AI: Fun hidden-information audio localization puzzle (human plays blindfolded catcher navigating by acoustic cues against AI swarm).
- Group games (n>4): Naturally supports 5 to 8 autonomous agents.
- Physical-skill games: Tactile touch and 3D sound spatialization are modeled via polar noise cones and deduction matrices.
- What is lost in the adaptation: The live physical laughter, dodging under arms, and the sensory disorientation of wearing a real cloth blindfold.
## Implementation hints
Low-to-moderate complexity. The core is an auditory perception filter and polar collision detector. Milestone 1: implement polar movement, boundary constraints, and noise-injected sound emitters; Milestone 2: build a particle-filter localization AI for the blindfolded catcher.
Download as JSON
Download the full game corpus entry for this game as structured JSON.
Download JSONCite this page
Kanamachi (Blind Bee) (Bangladesh). (n.d.). Games of Countries. https://gamesofcountries.com/games/bd-bangladesh/kanamachi
@misc{gamesofcountries-bd-bangladesh-kanamachi,
title = {Kanamachi (Blind Bee) (Bangladesh)},
year = {n.d.},
publisher = {Games of Countries},
url = {https://gamesofcountries.com/games/bd-bangladesh/kanamachi}
}