Imagine a position where it is our move and we can capture our opponent's last move (m).
m
capture m
If we search this subtree, and find that our opponent's best move is to play at "m" again:
m
Then capturing m must be a better move than any that don't capture m.
If there are two or more ways of capturing m, they should probably be compared.
I have seen one game against a strong human player where this technique would have won the game for the AI.
Imagine a position where it is our move and we can capture our opponent's last move (m).
If there are two or more ways of capturing m, they should probably be compared. I have seen one game against a strong human player where this technique would have won the game for the AI.