Deep learning applied to turn-based board games

Due to the astonishing growth rate in computational power, artificial intelligence is achieving milestones that were considered as inconceivable just a few decades ago. One of them is AlphaZero, an algorithm capable of reaching superhuman performance in chess, shogi and Go, with just a few hours of self-play and given no domain knowledge except the game rules. In this paper, we review the fundamentals, explain how the algorithm works, and develop our own version of it, capable of being executed on a personal computer. Despite the lack of available computational resources, we have managed to master less complex games such as Tic-Tac-Toe and Connect 4. To verify learning, we test our implementation against other strategies and analyze the results obtained