Re: [computer-go] Alpha-beta pruning vs transposition tables

[Don Dailey <drd@xxxxxxx>]

> The problem is that, when a branch is pruned, the root of that branch 
> is given an evaluation which may not be entirely accurate.  It's 
> enough to prove that the branch is irrelevant for this move, but 
> storing this value in a transposition table can lead to an incorrect 
> result when the position is encountered again.

>From this description, I think you have implemented the TT wrong.  The
score you store in the table in this case is a lower bound (if it was
due to a beta cutoff.)  You must store a flag to indicate this so that
you don't use it later.  When you see this value later in the table it
cannot be used directly to produce a score, it can only be used to
raise alpha (which may produce a cutoff which is good.)

I will be happy to review the relevant piece of code if you send it to
me.

- Don




_______________________________________________
computer-go mailing list
computer-go@xxxxxxxxxxxxxxxxx
http://www.computer-go.org/mailman/listinfo/computer-go/