For the second coursework of my Artificial Intelligence module I created a Reinforcement Learning agent that played a modified version of the popular children's game Connect Four. The game was cut down in both board size and win conditions (effectively making it Connect Three) in order to save on processing time. I trained my agent on 1000 games of Connect Three against a random opponent, 20 times. This produced an impressive result of an 80% win/loss ratio after the 20 rounds of training.
**Source code available on request**