> best female shogi player lost against a "computer". http://mdn.mainichi.jp/mdnnews/news/20101012p2a00m0na012000c.html
> The computer consisted of 3 or 4 different programs, > teamed up in some way. Maybe, Darren Cook can get > ideas from this for his bot-team which has to play > against John Tromp in December 2010 in London. I've already been pursuing this idea, as I think it is very promising. In addition to the unscientific evidence from my sm9 project, I was delighted to discover at CG2010 that there were two papers (by the same authors) on the subject of team play that give some rigorous results: Consultation Algorithm for computer Shogi: Move Decisions by Majority Optimistic Select Rule better than Majority Voting System Even 3-4 instances of the same program, each with some random noise added, is able to beat the same program. Once you have multiple programs with very different strengths, covering each other's weaknesses, the gain in strength can be very significant. Another advantage is it scales very naturally to a cluster. I have a script that runs for 9x9, controlling Fuego, Mogo, Gnugo, Valkyria and Many Faces, using GTP and TCP/IP sockets. But it needs work to be more stable, is untested on 19x19, and has no time control (it is designed so I can say: "talk among yourselves, take as long as you need, and tell me the best move in this situation"). So, I'm fairly pessimistic it could be used in December. Darren -- 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
