Building increasingly complex poker AIs: hand-evaluator based hand-evaluator plus CFR deep CFR neural net approach Then use these to build a poker trainer that looks like the lichess puzzles.