It works fine with komi 1.5 or 2.5 -- black wins and white wins, respectively. The situation below may be tricky to deal with. I'm guessing the problem is due to the bad move opening up a large area and extending the game, thus introducing uncertainty. And Pachi is preferring that uncertain outcome over a tie.
On Fri, Jul 15, 2011 at 7:19 PM, Michael Williams < [email protected]> wrote: > I found something else. > > This command results in white playing a very bad move which costs the game: > > twogtp.py --black '../pachi -t =100000' --white '../pachi -t =100000' > --size 4 --komi 2 > Eventually, this position occurs with white to play. > > A B C D > +---------+ > 4 | . X X . | > 3 | X O O O | > 2 | X X X O | > 1 | X . O O | > +---------+ > > Instead of passing and going for a tie with mutual life in seki, white > places his whole army in atari with a move at B1 and looses. > >
_______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
