Formal Specification
Blind man's buff (Finnish sensory tag) — Formal Specification
Finland · chase-tag · Confidence: high
← Back to game pageImplementation Suitability
Suitability: engine-readyFidelity: discretizedGap-fills: 3
# Sokko — AI-optimized specification
## General idea
A traditional sensory tag and identification game. A solitary blindfolded player (*sokko*) stands inside a bounded arena, unable to see, and must navigate using audio localization (footstep sounds, taunts, hand claps) and tactile reach to tag or identify one of the free evaders. Evaders tiptoe and tease around the blind catcher within the boundary. Once tagged or correctly identified by touch, roles swap.
## Formal model
- **Abstraction taken:** Continuous 3D acoustic propagation and human blindfolded movement are discretized into a 2D bounded grid . Vision for the Catcher is completely disabled (); instead, evaders emit sound cues upon moving or taunting, which the Catcher perceives as a noisy acoustic bearing vector . Tagging requires occupying an adjacent cell and executing a touch/identify action.
- **Player/agent count:** agents (), configured as 1 Blindfolded Catcher and Evaders.
- **Turn model:** Simultaneous discrete execution per tick.
## Formal rules
R1. **Arena & Setup:**
- A bounded 2D grid of size (e.g. ).
- All agents begin inside . One agent is designated `Sokko` (blind state: zero vision). All other agents are `Evaders`.
- The game runs across successive rounds.
R2. **Evader Actions & Acoustic Emission:**
- On each tick , each Evader chooses an action from:
- `move(dx, dy)`: where . Generates noise intensity .
- `creep(dx, dy)`: [gap-fill] Move 1 cell at half speed (takes 2 ticks). Generates low noise intensity .
- `taunt()`: Generates high noise intensity at current cell.
- `stand_still()`: Generates zero noise ().
- Out-of-bounds rule: If an evader moves outside , the evader is automatically captured and becomes the new `Sokko`.
R3. **Sokko Navigation & Perception:**
- `Sokko` chooses movement vector and/or `reach_tag(direction)`.
- `Sokko` receives an acoustic observation vector consisting of perceived sound bearings for each sound generated this tick within hearing radius .
R4. **Tagging & Role Transfer:**
- If `Sokko` executes `reach_tag(direction)` into a cell occupied by an Evader , is touched.
- **Standard Tag:** immediately becomes the new `Sokko`, the former `Sokko` becomes an `Evader`, and a new round begins.
- [gap-fill] **Identification Variant:** If playing the identification variant, `Sokko` submits an identity guess . If , role swap occurs; if , is released and receives a 5-tick head-start immunity.
R5. **Match Resolution:** [gap-fill] In a formal evaluation match, play proceeds for ticks. The agent with the minimum cumulative time spent in the `Sokko` role is declared the winner.
## Edge cases
- **Tie at the end condition:** If multiple agents have identical minimum time as Sokko, tie is broken by the fewest number of times being tagged.
- **A player with no legal move:** Stand still is always available.
- **Repetition / cycles:** Role swaps cycle continuously; match terminates strictly upon reaching the tick ceiling .
- **Simultaneous claims on the same resource or position:** Multiple evaders can occupy the same cell; if Sokko tags a multi-occupant cell, the agent with the lowest ID is tagged.
- **Player elimination mid-round; turn order repair:** Not applicable — no player elimination; roles rotate dynamically.
- **Resource exhaustion (pieces, tokens, space, time):** Spatial bounds and player count are invariant. Time bounded by .
- **Termination guarantee: why the game provably ends:** Guaranteed by the finite match horizon .
## State representation
- **Persistent state:** Grid bounds ; total players ; active Sokko agent ID; match tick counter ; cumulative ticks spent as Sokko per agent.
- **Per-tick ephemeral state:** Array of submitted actions; list of emitted acoustic events ; noisy acoustic observation vector delivered to Sokko.
- **Observable information:** Asymmetric / imperfect information. Evaders observe full visual state. Sokko observes zero vision and only noisy directional audio signals.
## Action space
- `move(dx, dy)` / `creep(dx, dy)` / `taunt()` / `stand_still()`: (Evaders).
- `move(dx, dy)` / `reach_tag(direction)` / `guess_identity(agent_id)`: (Sokko).
## Playability notes
- **Single-player-vs-AI:** Highly compelling as an imperfect-information stealth challenge. Human plays Sokko tracking AI evaders via audio radar pings, or human plays an evader sneaking around an AI blind predator.
- **Group games ():** Naturally scales to any group size ; can be reduced to 1 Sokko vs 1 Evader duel in a smaller grid with obstacles.
- **Physical-skill games:** Physical blindfolded proprioception and tactile identification are abstracted into acoustic sensor noise models and classification checks.
- **What is lost in the adaptation:** The physical disequilibrium of being spun around 3 times, the claustrophobia of the cloth blindfold, and the tactile feel of clothing and hair.
## Implementation hints
- Implement Sokko's observation as a Gaussian bearing sensor conditioned on emitted sound energy.
- Branching factor: movement options + 4 interaction types per agent.
- Suggested first milestone: Implement an 8×8 grid with 1 Sokko and 1 Evader, verifying that an evader's `move` produces an audio signal that allows Sokko to home in and execute `reach_tag`.
Download as JSON
Download the full game corpus entry for this game as structured JSON.
Download JSONCite this page
Blind man's buff (Finnish sensory tag) (Finland). (n.d.). Games of Countries. https://gamesofcountries.com/games/fi-finland/sokko
@misc{gamesofcountries-fi-finland-sokko,
title = {Blind man's buff (Finnish sensory tag) (Finland)},
year = {n.d.},
publisher = {Games of Countries},
url = {https://gamesofcountries.com/games/fi-finland/sokko}
}