Formal Specification
Cabbage Thief — Formal Specification
Estonia · chase-tag · Confidence: medium
← Back to game pageImplementation Suitability
Suitability: adaptation-requiredFidelity: discretizedGap-fills: 8
# Kapsavargamäng (Cabbage Thief) — AI-optimized specification
## General idea
A multi-phase dramatic roleplay and social contest game combining bluff-distraction mechanics, physical state signaling (stiff vs limp cabbage heads), blind team-partitioning, and a concluding strength tug-of-war. The thief attempts to harvest high-value cabbages while deceiving the gardener, but the final outcome is decided when the stolen cabbages are divided blindly between both leaders for a contest of collective pulling strength.
## Formal model
- **Abstraction taken:** The conversational comedy and physical neck-pressing test are replaced by discrete posture selection (`STIFF` vs `LIMP`), distraction success probabilities , and a stochastic aggregate tug-of-war resolver.
- **Player/agent count:** 1 Gardener agent (), 1 Thief agent (), Cabbage agents (, ).
- **Turn model:** Sequential phase-based execution across 3 distinct phases:
1. *Harvest Phase:* Alternating distraction and inspection rounds.
2. *Court Sorting Phase:* Sequential blind binary choice by stolen cabbages.
3. *Tug-of-War Phase:* Simultaneous aggregate force pull resolution.
## Formal rules
R1. Setup:
- Cabbages are placed in the Garden patch with initial physical posture state set to neutral.
- Gardener starts with Garden status `LOCKED`. Thief starts in the Thief Den.
R2. Phase 1 — Harvest & Inspection:
- Thief submits a distraction attempt. Success probability is evaluated as [gap-fill: modeled as a base probability decreasing by per repeated attempt].
- If distraction succeeds: Gardener looks away for 1 tick; Thief enters patch and selects 1 Cabbage to inspect.
- Inspection test: commits posture .
- If `STIFF` ("Hard Cabbage"): Thief steals to the Thief's Den ().
- If `LIMP` ("Rotten Cabbage"): Thief rejects ; remains in the patch ().
- If distraction fails: Thief is chased back to Den without inspecting.
- [gap-fill: Phase 1 concludes when all cabbages have been inspected or when the Thief has failed 3 consecutive distraction attempts].
R3. Phase 2 — Court Trial & Secret Word Sorting:
- Gardener confronts Thief in Court. Gardener secretly chooses code-word (e.g. "Apple"); Thief secretly chooses code-word (e.g. "Pear") from a public binary pair ().
- All stolen cabbages () plus any remaining garden cabbages are presented with the blind choice: Choose or .
- Each cabbage agent independently submits choice without knowing which leader holds which word.
- Agent mapping: Cabbages choosing join Team Gardener; cabbages choosing join Team Thief.
R4. Phase 3 — Tug-of-War Resolution:
- Team Gardener (Gardener + cabbages) pulls against Team Thief (Thief + cabbages).
- Each participant has a strength attribute [gap-fill: base pull strength ].
- Total team force: , , where .
R5. Win/Loss Condition:
- If , Gardener wins the dispute and reclaims all produce.
- If , Thief wins the dispute and keeps the harvest.
- [gap-fill: If , a sudden-death single-cabbage runoff pull is executed; if still tied, victory goes to the side with fewer members].
## Edge cases
- Tie at the end condition: Resolved via single-member runoff or team size tiebreaker in R5 [gap-fill: explicit tie-break rule].
- A player with no legal move: Not applicable; agents choose from enumerated discrete posture, word, or pull actions.
- Repetition / cycles: Phase 1 has a strictly finite cabbage count and a 3-consecutive-failure cap [gap-fill: cycle bound in Phase 1].
- Simultaneous claims on the same resource or position: In Phase 2, Gardener and Thief select words from ; if both submit the same word, Gardener takes priority and Thief receives the alternate [gap-fill: word conflict resolution].
- Player elimination mid-round; turn order repair: Rotten cabbages not stolen in Phase 1 participate in the final trial sorting to prevent idle agent starvation [gap-fill: inactive cabbage reintegration].
- Resource exhaustion: All cabbages are partitioned into exactly two non-empty or empty teams; no tokens are lost.
- Termination guarantee: Phase 1 is bounded by inspections and failure caps; Phase 2 is an exact -step binary map; Phase 3 is a single-step stochastic force evaluation. Total execution is strictly finite.
## State representation
- Persistent state:
- `cabbage_status`: Map of Cabbage ID .
- `stolen_count`: Integer .
- `gardener_word`, `thief_word`: Binary assignment from .
- `team_gardener_members`: List of Agent IDs.
- `team_thief_members`: List of Agent IDs.
- Per-phase ephemeral state:
- Phase 1: `consecutive_distraction_fails`, `current_cabbage_posture`.
- Phase 3: `pull_forces` .
- Observable information:
- Phase 1: Posture is revealed to Thief only upon inspection; distraction success is public.
- Phase 2: Word-to-Leader mapping () is hidden from Cabbage agents until choices are committed.
## Action space
- `attempt_distraction(novelty_val)`: (Thief) Trigger distraction check.
- `inspect_cabbage(cabbage_id)`: (Thief) Inspect target.
- `set_posture(posture)`: (Cabbage) Choose `STIFF` or `LIMP`.
- `assign_secret_word(word)`: (Gardener/Thief) Choose or .
- `choose_word(word)`: (Cabbage) Select or blindly.
- `exert_pull(intensity)`: (All agents) Contribute force in Phase 3.
## Playability notes
- Single-player-vs-AI: Human can play Thief trying to optimize cabbage harvesting and word selection, or human plays Gardener defending the patch.
- Group games (n>4): Reduces to 1 Gardener vs 1 Thief with 2 Cabbages without structural distortion, though larger cabbage groups enhance tug-of-war tension.
- Physical-skill games: Physical body tension (neck rigidity) is modeled as a game-theoretic choice; tug-of-war pulling is modeled via normal strength distributions.
- What is lost in the adaptation: The theatrical humor of folk excuses, improvised comedic accents, and physical tactile joy of hand-on-head inspection.
## Implementation hints
Treat the three phases as a finite state machine: `HARVEST -> COURT_SORTING -> TUG_OF_WAR -> TERMINAL`. First milestone: Implement Phase 2 word-assignment and Phase 3 force resolution before building Phase 1 distraction logic.
Download as JSON
Download the full game corpus entry for this game as structured JSON.
Download JSONCite this page
Cabbage Thief (Estonia). (n.d.). Games of Countries. https://gamesofcountries.com/games/ee-estonia/kapsavargamang
@misc{gamesofcountries-ee-estonia-kapsavargamang,
title = {Cabbage Thief (Estonia)},
year = {n.d.},
publisher = {Games of Countries},
url = {https://gamesofcountries.com/games/ee-estonia/kapsavargamang}
}