Formal Specification
Indonesian Hopscotch / Sunda Manda — Formal Specification
Indonesia · dexterity-throw · Confidence: low
← Back to game pageImplementation Suitability
Suitability: engine-readyFidelity: probabilistic-proxyGap-fills: 5
# Engklek — AI-optimized specification
## General idea
A turn-based grid navigation and territory acquisition game played on an 8-cell topological graph. Agents throw a marker into progressive target cells, hop through valid unoccupied path nodes while avoiding owned territory cells, and earn private "sawah" territory nodes via blind terminal throws. The board topology dynamically constrains path traversal as more nodes are claimed.
## Formal model
- **Abstraction taken:** Physical throwing precision and one-legged balance are abstracted into success probabilities and . Court geometry is modeled as a connected graph of 8 cells: .
- **Player/agent count:** 2 to 5 agents in a round-robin rotation.
- **Turn model:** Strictly sequential.
## Formal rules
R1. Graph structure: 8 cells with adjacency list:
- .
- Cells and are paired wing nodes allowing simultaneous two-foot rest.
R2. Progress state: Each agent tracks `current_target_cell` and `owned_sawah` (set of claimed cell IDs).
R3. Turn action:
- Active agent throws marker to `current_target_cell`. Success probability [gap-fill].
- If throw fails (lands out of bounds/on line), turn ends immediately.
- If throw succeeds, agent traverses the path from to and back to :
- Must hop over the cell containing the marker.
- Must hop over any cell .
- If a valid jump sequence across adjacent un-owned nodes exists, balance check succeeds with [gap-fill].
- If balance check fails or path is obstructed by impassable consecutive opponent sawahs, turn ends.
- Upon returning safely, `current_target_cell` increments by 1.
R4. Territory claiming (*Sawah*):
- When an agent completes `current_target_cell == 8`, they perform a blind backward throw.
- A random unclaimed cell is selected with uniform probability [gap-fill].
- is added to `owned_sawah` of that agent. `current_target_cell` resets to 1 to begin a new lap.
R5. Win condition: The game ends when all 8 cells are claimed as sawah (or max turns is reached). The agent with the most claimed sawah cells wins.
## Edge cases
- Tie at the end condition: If two agents claim an equal number of sawah cells when all cells are exhausted, the agent who completed the progression in fewer total turns wins.
- A player with no legal move: If an agent's path is blocked by impassable consecutive opponent sawahs, they forfeit their turn without progress.
- Repetition / cycles: Bounded by max turn horizon and finite available sawah cells ().
- Simultaneous claims on the same resource or position: Not applicable; turns and cell claims are strictly sequential.
- Player elimination mid-round: If an agent forfeits, their claimed sawah cells become neutral (passable by all) and turn rotation skips them [gap-fill].
- Resource exhaustion: When all 8 cells are claimed, game immediately triggers terminal score tally.
- Termination guarantee: Enforced by the finite pool of 8 claimable cells and hard upper turn limit .
## State representation
- Persistent state:
- `agents`: array of objects `{id, target_cell, sawah_cells: []}`.
- `claimed_cells`: mapping of `cell_id -> owner_id`.
- `active_agent_idx`: integer .
- `turn_count`: integer.
- Per-turn ephemeral state:
- `marker_pos`: integer .
- Observable information: Fully observable (all positions, targets, and sawah ownerships are public).
## Action space
- `execute_hopscotch_lap()`: Automated trajectory execution evaluating throw accuracy and traversal checks.
- [gap-fill] choose_hop_path(route)`: For branched cells (e.g. vs ), choose foot placement node.
## Playability notes
- Single-player-vs-AI: Highly playable as a spatial route planning challenge against heuristic/RL bots.
- Group games (n>4): Reduces naturally to 2 agents (1v1) without structural changes.
- Physical-skill games: Physical balance and motor precision are modeled via distance-decay probability functions.
- What is lost in the adaptation: The somatic exertion of maintaining balance on one leg while bending over to pick up a stone is abstracted.
## Implementation hints
Graph navigation is easily modeled with Dijkstra/BFS pathfinding to verify hop legality between available non-hostile nodes. First milestone: Implement 1v1 on an 8-cell array with throw success checks and sawah ownership locks.
Download as JSON
Download the full game corpus entry for this game as structured JSON.
Download JSONCite this page
Indonesian Hopscotch / Sunda Manda (Indonesia). (n.d.). Games of Countries. https://gamesofcountries.com/games/id-indonesia/engklek
@misc{gamesofcountries-id-indonesia-engklek,
title = {Indonesian Hopscotch / Sunda Manda (Indonesia)},
year = {n.d.},
publisher = {Games of Countries},
url = {https://gamesofcountries.com/games/id-indonesia/engklek}
}