https://sirlis.github.io/posts/reinforcement-learning-Temporal-Differences/