I'm currently concentrating on the opening. I can now build a UCT tree of about 14 billion nodes before I run out of storage. Concentrating so much on one
position only makes sense where it can be reused, which is at the start of the game. On 7x7, creating that many nodes takes about a day. I still haven't built
it into a playing engine though, so I have no idea how much it helps or how good the tree coverage is in a real game.
Michael Alford wrote:
delurk/
I have a bet with Frank de Groot, that bet is that he cannot write a
program that can beat me in an even game before I die (I am currently 64
yrs, the bet is several year old), this came about from his insisting
that the game is a FSM, which may be true in theory, the problem is the
word "finite". My current playing strength is 2d on IGS, this post is
inspired partly by recent comments by Yamoto, so:
The game of Go has four stages, joban, chuban, o-yose, and yose. I
expect computer programs to be very good at yose (not counting some of
the throw-in stuff on first line), pretty good at o-yose, good at
chuban, and utterly worthless in joban. Now, in joban you play fuseki
patterns, said patterns have a definite purpose, kobayashi, high
chinese, low chinese, ni ren sei, san ren sei, mini-chinese, etc, and
yet I read posts here debunking the value of opening books. I think this
is a very serious flaw in your approach. re Yamoto, all programs do yomi
very well, none of them play with kankaku, so I think that the next step
in your developing algorithms for this game *has* to include opening
books, otherwise I see no way for the programs to play with "kankaku".
my $0.02
lurk/
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
