On Tue, May 13, 2008 at 01:34:37PM -0400, Don Dailey wrote: > The point after an illegal move is quite a bit more likely to be > selected. If the list had just 1 illegal point, then the point after > it in the list is twice as likely to be selected as any other point. > Perhaps if you added a little noise after each move (perhaps swapping to > points at random) you could minimize any seriously bad effects?
I think the problem becomes more pronounced near the end of the simulation, when most of the boars is filled with stones or 1-point eyes. If, at that time, there are two legal moves left, one near the beginning of the list, and one near the end, then the second one can end up as much more likely to be chosen. And if the first one is the key point to the position, it may never even be noticed. No idea how likely this scenario is in real games, but at least it is possible. And if your search tree is large enough, once-in-a-thousand situations occur all the time... -H -- Heikki Levanto "In Murphy We Turst" heikki (at) lsd (dot) dk _______________________________________________ computer-go mailing list computer-go@computer-go.org http://www.computer-go.org/mailman/listinfo/computer-go/
