A developer built a greedy solver for the puzzle game Block Blast that scored moves by immediate line clears. The bot chose the highest short‑term reward, often stacking blocks and creating impossible states a few turns later. This short‑sightedness made it win battles but lose the war in future plays.

To fix this, the author replaced the naive score with a weighted heuristic that evaluates future flexibility, combo potential, danger detection, and accessibility. Each factor receives a weight—0.4, 0.3, 0.2, 0.1—so the solver balances immediate clears against long‑term board health. The new function consistently outperforms the greedy baseline in simulation tests daily.

The upgraded solver survives 30‑40% longer, raising average scores and producing moves that feel intentional rather than reactive. The next milestone is a minimax algorithm with pruning, allowing the bot to look two to four turns ahead. Mastering this could unlock optimal play and reshape how we approach puzzle AI.