I ran a 2 hour test where my compute played a bunch of poker with itself. I played 100 strategies for 1st position (for those of you who know poker) and 100 strategies for 2nd position against eachother. Every 1st position strategy played every 2nd position strategy 10,000 hands. Then the best one for each strategy (the one who had the most winnings after playing 1,000,000 hands) is bread into 100 more AI. The breeding used a random binomial to determine the percent it should deviate from a strategy point in the positive or negative direction. (-1 would result in a 0 for that point and 1 would result in a 1.) Anyways, here are the pretty results (p.s. bet size = 1, pot size = 1, 1 round of betting, no draws etc, 1 bet cap. A player bets hands outside of the range, and checks hands inside the range. A player player will call with hands better than or equal to the middle value (better means less), and fold worse hands. Hands are random values between 0 and 1.)