> The version of Valkyria playing on LG is not the same as tested on CGOS. > I have done some things to try to overcome the noise problem of MCTS and > memory limitations. The games on LG is played with up 12-24 hours using > 2 threads (several positions are searched in parallel as games are > played in parallel on LG) on a 4 core machine. > > This version of Valkyria is currently undefeated (there have been losses > in the past due to operating mistake and with short times and the > standard search).
To give an idea of Valkyria's strength at slow 9x9 games: http://www.littlegolem.net/jsp/tournament/tournament.jsp?trnid=go9.ch.25.1.1 All players here are very strong *and* play a lot of dan-level 9x9 go. (E.g. Nao is 7-dan, and has won this 9x9 tournament three times.) (Gerhard and sm9 are human-computer experiments; LEESD may be a program, his account info is confusing [1].) This was the worst result (5-3 for joint 3rd) for sm9 in 4 years, but there has been no loss in strength [2]. The difference from previous tournaments was people are not making a mistake after move 10 any more. Or, at least, that is how it felt. I'm very impressed that Valkyria didn't have this problem. It started the tournament at 1 kyu, so I suppose it is possible that it took opponents by surprise. But I doubt that. Anyway, we'll see next tournament :-) Darren [1]: http://www.littlegolem.net/jsp/info/player.jsp?plid=21689 [2]: But also no gain in strength, and the planned improved opening book is still in limbo. Sadly I don't think that situation will change this year. -- Darren Cook, Software Researcher/Developer 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
