No long ago I was considering a scheme something like "bias uct formula by
solid points" to cure the problem of underestimation of the edge and
corner territory, but still have no time to try. I believe it is one of the
solutions to the notorious problem that MCTS programs usually like center
territory more.
Take the empty position as an example: a center move A and a corner move B
might both have 47% winning rate, but B should have higher score in UCT
formula computation. The reason is simple: from the ownership information of
the playouts, there is no solid points for A (everywhere both sides own
almost 50%) while B has some solid points at the corner (such as Black owns
that corner points in 60% of the playouts).
Aja
-----原始郵件-----
From: Yamato
Sent: Sunday, January 08, 2012 8:30 PM
To: [email protected]
Subject: Re: [Computer-go] replacing dynamic komi with a scoring function
(2012/01/09 9:56), Jouni Valkonen wrote:
There is indeed a problem with Dynamic komi with Zen. Zen often loses
the handicap games if black tries to minimize the move count. Often if
it is possible to bring game to small yose in around move 180 or so and
if not too much behind, then Zen most likely will lose. I have played
few games where Zen noted only when filling last dames that it is losing
the game by few or half points and then resign. Although, one game was
that I lost by ½ points, because I accidentally defended unnecessarily
instead of taking the last dame. One game was that i was about ten
points behind around move 180, but then Zen played a slack small yose,
and lost by 2½ points. Also good and very easy strategy against zen in
handicap games is to take all the sides and give center territory to the
Zen. Zen almost always will take the center territory as too small and
gives sides as too big.
I think people often confuse the evaluation problem and the dynamic komi
problem. A more urgent problem is the underestimation of the edge and
corner territory.
--
Yamato
_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
