Engine: policy gradient learning, 84000 self-play games (chapter 9)
new game
Black to play