Multi-Armed Bandits for <i>Minesweeper</i>: Profiting From Exploration–Exploitation Synergy

A popular computer puzzle, the game of Minesweeper requires its human players to have a mix both luck and strategy succeed. Analyzing these aspects more formally, in our research, we assessed feasibility novel methodology based on reinforcement learning as an adequate approach tackle problem presented by this game. For purpose, employed multi-armed bandit algorithms which were carefully adapted order enable their use define autonomous computational players, targeting make best some peculiarities. After experimental evaluation, results showed that was indeed successful, especially smaller boards, such standard beginner level. Despite fact, main contribution work is detailed examination from perspective, led various original insights are thoroughly discussed.