Re: [computer-go] Modern brute force search (Gods handicap)

["chrilly" <chrilly@xxxxxxxx>]

>
>>It sounds like GO isn't as complicated as I thought it was, due to the
>>fact that many of you believe masters are pretty close to optimal play.
>
Top-chess players can of course also calculate deeper than 12 Plies. But
this
12 Plies are the calculation depth of "intuition". When playing against a
perfect (or tactical strong) opponent, one must not only check the forced
lines, but also the lines without direct threats. According my experience in
playing against Top-GMs there is indeed such a 12-Ply barrier.

The direct consequence of this barrier is: Human chessplayers are very far
away from optimal play. God must just create quiet threats >12 Plies.

If Go-Players state they can read easily to 30 Plies it does not mean that
their barrier is 30 Plies. Once they are reading, they have already realized
that there is some tactic around. The essential point is: Do they read any
30 Ply threat? I canīt believe it.

But even 30 Plies sounds not impossible.If one would make a project like
Deep-Blue it would be within 10 years at least within reach.
As a first guess the effective branching factor is somewhat less than half
of the theoretical optimal branching factor => 7-8.
Modern brute-force  programms do not search brute-force. The Null-Move and
other pruning techniques are very efficient. In case of Go the programm must
be clever enough to realize Sente. Otherwise one can not use the Null-Move.
This is probably a very big IF for using the Null-Move.

Chrilly



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