Yes, I implemented this and it seems to play better in handicap games vs humans, but no rigorous testing.
> -----Original Message----- > From: [email protected] [mailto:computer-go- > [email protected]] On Behalf Of Darren Cook > Sent: Sunday, October 03, 2010 4:01 PM > To: [email protected] > Subject: Re: [Computer-go] Exhibition match > > > For high handicap games taking black, giving white a big initial komi that > > decreases with time seems to help a lot. ... > > Does that mean you've implemented and shown some improvement in Many > Faces? (In handicap games against humans?) (As both black and white?) > > I think Petr Baudis has shown a positive benefit in even games too, but > hasn't published those results yet. > > And I think Magnus (Valkyria) got a similar result (though not sure if > that was just in handicap games). > > Darren > > > > Otherwise at the first move, the > > program thinks it has over 80% win rate, and plays passively. This is not > > the way to win against a stronger player. > > > > > Magnus, Erica does not have dynamic komi. I didn't expect to win, so I > never > > tested any handicaped game before. Also, I changed the komi 7.5 to 0.5 > just > > before the game start. I believe Zen, Many Faces, even Mogo, must do much > > better in handicaped games. > > -- > Darren Cook, Software Researcher/Developer > > http://dcook.org/gobet/ (Shodan Go Bet - who will win?) > http://dcook.org/work/ (About me and my work) > http://dcook.org/blogs.html (My blogs and articles) > _______________________________________________ > 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
