10-04-2020, 11:15 PM
(This post was last modified: 10-04-2020, 11:17 PM by AlekErickson.)
Normally , I would drop this, but I want to squeeze every drop of AI strength out of this implementation.
So here goes:
Actually, the difference is not necessarily just one pass.
First of all, in this game, under the current ruleset, when one player runs out of moves, the other player might have hundreds of moves left, each with a large branching factor. So, even though no action taken by either player can change the game result, the game tree still has a significant complexity remaining.
In the case when the AI is assessing the position at the end of a MCTS play-out, and its play-out has reached a position where it has run out of moves, will it always understand that such a "certainly lost game", is actually certainly lost?
If not, then we would help the AI strength by reinforcing the point that the game is over when a player has run out of moves.
So here goes:
Actually, the difference is not necessarily just one pass.
First of all, in this game, under the current ruleset, when one player runs out of moves, the other player might have hundreds of moves left, each with a large branching factor. So, even though no action taken by either player can change the game result, the game tree still has a significant complexity remaining.
In the case when the AI is assessing the position at the end of a MCTS play-out, and its play-out has reached a position where it has run out of moves, will it always understand that such a "certainly lost game", is actually certainly lost?
If not, then we would help the AI strength by reinforcing the point that the game is over when a player has run out of moves.