Formal Specification
Moroccan knucklebones (Kaab) — Formal Specification
Morocco · dexterity-small · Confidence: medium
← Back to game pageImplementation Suitability
Suitability: engine-readyFidelity: probabilistic-proxyGap-fills: 1
# Tabba — AI-optimized specification
## General idea
A multi-level physical dexterity jacks/knucklebones challenge. The active player tosses a marker bone into the air and must execute precise subset groupings from a ground cluster of 4 astragalus bones (singles, pairs, triplets, quad, and bridge sweep) and catch the falling marker before impact. Faults pass the turn to the next player.
## Formal model
- **Abstraction taken:** The 3D juggling and hand-manipulation physics of sheep astragali are modeled as a discrete level sequence with stochastic execution success: .
- **Player/agent count:** 2 to 6 agents in fixed rotational order.
- **Turn model:** Sequential turns. A player continues taking attempts upon success and yields the turn upon a fault.
## Formal rules
R1. **Initial State & Levels:**
- 5 physical knucklebones: 1 airborne tosser () and 4 ground bones ().
- Each player starts at `level = 1`, `completed = false`.
- Levels:
- Level 1: Scoop 4 singles ().
- Level 2: Scoop 2 pairs ().
- Level 3: Scoop triplet + single ().
- Level 4: Scoop all 4 together ().
- Level 5 (*Bridge*): Form tunnel with non-dominant hand, sweep all 4 bones through tunnel in 1 toss.
R2. **Attempt Execution & Success Probabilities:**
- On each level attempt, the agent executes `toss_and_gather(level, subset_grouping)`:
- Ground bones are distributed randomly on the 2D plane: .
- Ground dispersion .
- Probability of clean scoop and catch: .
R3. **Faults & State Persistence:**
- If check succeeds: Current sub-step is cleared. If all sub-steps of the level are cleared, `level \gets \text{level} + 1$, and player attempts next level.
- If fails: Player suffers a **Fault** (`status = fault`). Turn immediately passes to next player in rotation. Player's progress remains saved at `current_level`.
R4. **Victory Condition:**
- The first player to successfully complete Level 5 (*The Bridge*) wins the match.
## Edge cases
- **Tie at the end condition:** Not applicable — win triggers dynamically on an individual player's turn.
- **A player with no legal move:** Attempting the current level's gathering action is always legal.
- **Repetition / cycles:** Bounded by strictly non-decreasing level state and positive success probability . [gap-fill] If 40 aggregate rotation cycles pass without a winner, the player with the highest `level` progress wins.
- **Simultaneous claims on the same resource or position:** Turns are strictly sequential; board is reset per player's turn.
- **Player elimination mid-round; turn order repair:** If an agent drops out, rotation skips their turn index.
- **Resource exhaustion:** Exactly 5 knucklebones.
- **Termination guarantee:** Guaranteed by non-decreasing level progression and the 40-cycle cap.
## State representation
- **Persistent state:** Array of player levels `[L_1, ..., L_N]`, active player index , total cycle count.
- **Per-attempt state:** Coordinates of 4 ground bones , current sub-step index.
- **Observable information:** Fully observable state.
## Action space
- `toss_and_gather(level, subset_indices)`:
- `subset_indices`: List of ground bone IDs to scoop during current airborne toss window.
## Playability notes
- **Single-player-vs-AI:** Natural as a solitaire score-attack (completing all 5 levels with fewest faults) or competing against simulated AI players.
- **Group games ():** Standard round-robin rotation.
- **Physical-skill games:** Finger dexterity and hand-eye coordination abstracted into parameterized level failure rates.
- **What is lost in the adaptation:** The sensory clatter of real sheep astragali clicking together in a child's hand and the distinct yellow saffron dye scent.
## Implementation hints
- Implement as a 5-state Markov decision process with distance-dependent transition probabilities.
- First milestone: Implement levels 1–4 single-player state machine with deterministic inputs to verify level-advance triggers.
Download as JSON
Download the full game corpus entry for this game as structured JSON.
Download JSONCite this page
Moroccan knucklebones (Kaab) (Morocco). (n.d.). Games of Countries. https://gamesofcountries.com/games/ma-morocco/tabba
@misc{gamesofcountries-ma-morocco-tabba,
title = {Moroccan knucklebones (Kaab) (Morocco)},
year = {n.d.},
publisher = {Games of Countries},
url = {https://gamesofcountries.com/games/ma-morocco/tabba}
}