Home-brewed AI, not a NN. Similar in some ways to AlphaGo Zero's MCTS. Learns to play with 0 knowledge. No training data. No rollouts. It squeezes a lot of inf

Home-brewed AI, not a NN. Similar in some ways to AlphaGo Zero's MCTS. Learns to play with 0 knowledge. No training data. No rollouts. It squeezes a lot of info from each hand, more than a NN, which is why it needs less trials. Currently 100k hands and most respectable play.