Minesweeper Constraint Solver

Summary

Pairs CSP reasoning with heuristic pruning so every move can be traced and explained on randomized boards.

Problem

Solving Minesweeper reliably often turns into opaque brute force; operators need transparent reasoning about each move.

Approach

Tracks constraint propagation, logs decision traces, and prunes search branches while recording why certain cells stay unknown.

Highlights

• • Stored board state + propagation steps so move traces can be replayed.
• • Optional CLI decision-trace mode prints inference steps for debugging.
• • Seeded tournament benchmarking quantifies win-rate across controlled boards.

Results

In progress: adding benchmarks and visuals.

Trade-offs

• • The decision trace increases runtime slightly but keeps the workflow auditable for QA.
• • Deterministic seeding gives repeatability at the cost of covering less diverse board setups without additional permutations.

Next improvements

• • Add cached inference paths when similar board patterns reoccur to reduce repeated computation.
• • Surface how often human intervention is needed and why.

Links

• Repo