On 22, Nov 2007, at 9:35 AM, Don Dailey wrote:
This is one of many things in life that people refuse to believe -
regardless of the evidence. ...
Instead, people focused on highly selective searches. In order to
play strong it was clear that computers would have to look 20 or 30
ply
ahead
Our experience with SlugGo is that playing strength peaked only
a few ply deep in global lookahead. Our conclusion was that for
deeper lookahead the evaluation function that was used at the
leaf nodes became irrelevant because the real game was off in a
different branch than any of our expected paths. I would call our
tree "somewhat selective" near the top, but increasingly selective
deeper into the tree, so I am not sure how our results map onto
your argument.
If your argument is that it is worth exploring the entire tree to
whatever depth is possible, then our results are probably not
really evidence to the contrary.
Cheers,
David
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
