I feel like reinforcement learning fits best in the context of games from what I've ready so far ( in the context of deep learning, not counting other game AI techniques ).