Brainstorming

Machine Learning

Of some kind, obviously. Currently it's a very basic decision tree.
Possible avenues are looking at figuring out a reward point system and then running some simulations using Q-Learning.