Don Dailey wrote:
I'm thinking that if we estimate Aoba at 10d amateur and CrazyStone wins
with 8 stone handicap, it is roughly equivalent to beating a 2d player
without handicap and that we can subtract 2 stones to say that with
pretty high confidence CrazyStone is playing at least 1 kyu (but that's
it's much more likely Crazy Stone is stronger than this - after all it
performed in this one game at least as well as 2d player.)
Hi Don,
Crazy Stone played a few games on KGS, and has a strong 1k rank there.
When watching the games, I feel that its level is sometimes very high,
sometimes very low. If there is a semeai on the board that the playouts
evaluate wrongly, then its playing style is absolutely horrible (much
lower than 1k). Otherwise, it plays very strong (stronger than 1k).
Well, I am only 5k, so I cannot really tell. But when I see the horrors
it plays in some games, I suppose it must play much stronger than 1k in
some other games in order to get a rating of 1k.
Look for instance at these two games:
a win: http://files.gokgs.com/games/2008/8/23/CrazyStone-mandelbrot.sgf
a loss: http://files.gokgs.com/games/2008/8/23/CrazyStone-beoren.sgf
(with comments of the opponents at the end)
I have the feeling that MoGo has the same problem. It seems to be
typical. Against Kim Myung Wan, it was very clear. In the games it lost,
there were big semeais on the board. And MoGo played horribly. In my
opinion, this is what explains the apparent difference of level between
the blitz and the slow game. Not the time control. Of course, this is
just a guess. Maybe members of the MoGo team can confirm/infirm this.
When the playouts evaluate a critical semeai the wrong way, then no
supercomputer can help, even at long time control. Semeais require a
better algorithm, because no computing power can search them out with a
tree, and playouts have to be extremely intelligent in order to evaluate
them correctly.
Rémi
_______________________________________________
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
