maxpumperla / deep_learning_and_the_game_of_go

Code and other material for the book "Deep Learning and the Game of Go"
https://www.manning.com/books/deep-learning-and-the-game-of-go
953 stars 387 forks source link

If greedy_move is illegal, doesn't this implementation just pass? #43

Open artus-LYTiQ opened 4 years ago

artus-LYTiQ commented 4 years ago

https://github.com/maxpumperla/deep_learning_and_the_game_of_go/blob/c1add1ff272f8927a82d81c9ee430f9c13e062ef/code/dlgo/agent/alphago.py#L153

Heyho, where am I mistaken here: to me it looks like if the greedy_move is not legal, this loop will then simply try to select the same move (unsuccessfully) again till rollout_limit is reached. And then it will effectively directly call game_state.winner(). What did I miss?