Closed danzel closed 6 years ago
Implemented.
Current results (With depth 3 vs 1_6 exhaustive): https://gist.github.com/danzel/3ab37feaf357228c65d5ec855aef415d
Run 4 and 45 are not draws, but one AI wins each, so overall draw!
Took a look at the placement differences compared to 1_6, I think the mistakes are now that MCTS is suggesting we are going to buy pieces we aren't going to be able to place.
ExhaustiveMostFuturePlacements currently performs better than MCTSPreplace with the Tight placement preplacer.
We should be able to use MCTSPreplace with Exhaustive as the preplacer, which might perform better than both?