Hi, Thanks Remi for posting detail result. I add some info about this match.
Crazy Stone ran on Quad Core (3.0Hz) x2 Mac, and used 6 core out of 8 core. Remi said its speed was 30,000 playout/s on it. (at least in first game) KCConGUI is a commercial version of "Ginsei Igo 7" that is sold in Japan. http://www.silverstar.co.jp/02products/gigo7/gigo7.htm (in Japanese) It is latest version of KCC Igo. (Ginsei Igo 8 will be sold next month.) KCC Igo have won Gifu Challenge 4 times in a row. KCC Igo ran on Opteron 2.6GHz. Level is "Maximum". Time setting is "no limit". Play style is "Normal". Komi is 7.5. I updated latest update patch. Match was played on KGS. Time setting is 30 minutes and 10 seconds. (if program uses 30 minutes, it must play within 10 seconds.) But KCC does not have fine time control mode. (4 modes, "no limit", "long", "normal" and "blitz".) But it usually plays all its moves within 30 minutes on Opteron 2.6GHz. (KCC lost on time only one game.) Rule is Chinese. KCC does not have Chinese rule mode, so it played in Japanese rule. So after filling dame, endgame was a bit strange. I made GUI to GTP program. It looks work until KCC passes. After KCC passes, it often failed screen recognition. Regards, Hiroshi Yamashita ----- Original Message ----- From: "Rémi Coulom" <[EMAIL PROTECTED]> To: "computer-go" <[email protected]> Sent: Saturday, October 13, 2007 4:09 AM Subject: Re: [computer-go] best approach forward > Ian Osgood wrote: > > > > On Oct 11, 2007, at 11:01 AM, Rémi Coulom wrote: > > > >> In case nobody noticed, Crazy Stone won a match against KCC Igo this > >> summer, with 15 wins and 4 losses. The match was organized by Hiroshi > >> Yamashita. The games can be found in the KGS archives. > >> http://www.gokgs.com/gameArchives.jsp?user=kccongui&year=2007&month=7 > >> > >> Rémi > > > > Are you certain of the result? Two of the games I examined have scores > > that don't take into account removal of dead stones. (W+66.5 should > > be B+20, and W+48.5 looks like a 1 pt game.) This makes me wonder if > > some of the other losses and unfinished games (by score or time) are > > actually wins for KCC Igo on the board. I would certainly like to see > > the twenty games of this match validated, corrected, commented, and > > preserved for posterity. > > > > If this is valid, it is quite an achievement! I've updated my own > > ranking estimates accordingly: > > > > http://senseis.xmp.net/?GoPlayingPrograms%2FDiscussion#toc8 > > > > Ian > > This match was played through a system by Hiroshi Yamashita that > automatically reads the screen and sends mouse events, so that any GUI > program can be connected to GTP. This system seems to be not 100% > reliable, and does not handle any scoring protocol. > > This is quoted from a mail by Hiroshi Yamashita to me after the match: > > " > KCC and Crazy Stone played 20 games. > > I listed as follows. > > Crazy Stone has 14 wins and 5 losses. > One games stopped halfway, so I don't count this. > And I counted these two games to KCC win. > > It failed to recognize stones. 323th black move was not real move. > 1 http://files.gokgs.com/games/2007/7/22/CrazyStone-KCConGUI-6.sgf > KCC lost on time. But I think if KCC had time control, KCC could win. > 211th move was K-2. (it did not play) > 1 http://files.gokgs.com/games/2007/7/22/CrazyStone-KCConGUI-5.sgf > > > 0 http://files.gokgs.com/games/2007/7/22/CrazyStone-KCConGUI-10.sgf > ? http://files.gokgs.com/games/2007/7/22/CrazyStone-KCConGUI-9.sgf > 0 http://files.gokgs.com/games/2007/7/22/CrazyStone-KCConGUI-8.sgf > 1 http://files.gokgs.com/games/2007/7/22/KCConGUI-CrazyStone-2.sgf > 0 http://files.gokgs.com/games/2007/7/22/KCConGUI-CrazyStone.sgf > 0 http://files.gokgs.com/games/2007/7/22/CrazyStone-KCConGUI-7.sgf > 1 http://files.gokgs.com/games/2007/7/22/CrazyStone-KCConGUI-6.sgf > 1 http://files.gokgs.com/games/2007/7/22/CrazyStone-KCConGUI-5.sgf > 0 http://files.gokgs.com/games/2007/7/22/CrazyStone-KCConGUI-4.sgf > 0 http://files.gokgs.com/games/2007/7/22/CrazyStone-KCConGUI-3.sgf > 0 http://files.gokgs.com/games/2007/7/22/CrazyStone-KCConGUI-2.sgf > 0 http://files.gokgs.com/games/2007/7/22/CrazyStone-KCConGUI.sgf > 0 http://files.gokgs.com/games/2007/7/21/KCConGUI-CrazyStone-6.sgf > 0 http://files.gokgs.com/games/2007/7/21/CrazyStone-KCConGUI-2.sgf > 0 http://files.gokgs.com/games/2007/7/21/KCConGUI-CrazyStone-4.sgf > 0 http://files.gokgs.com/games/2007/7/21/KCConGUI-CrazyStone-3.sgf > 0 http://files.gokgs.com/games/2007/7/21/KCConGUI-CrazyStone-2.sgf > 1 http://files.gokgs.com/games/2007/7/21/CrazyStone-KCConGUI.sgf > 0 http://files.gokgs.com/games/2007/7/21/KCConGUI-CrazyStone.sgf > 1 http://files.gokgs.com/games/2007/7/18/CrazyStone-KCConGUI.sgf > > 5-14 > > I also attached Gnugo vs KCC result. > > GnuGo is 3.7.10. > Ginsei Igo 7 is a latest version of KCC (Dec 2006) > Ginsei Igo 5 is 2004 version of KCC. > > Ginsei Igo 7 is a same version that played against Crazy Stone. > > wins losses wining rate average score > GnuGo (Level 10) - Ginsei Igo 7, 37 142 0.21 -31.6 points > GnuGo (Level 10) - Ginsei Igo 5, 11 55 0.17 -31.3 > GnuGo (Level 16) - Ginsei Igo 5, 26 75 0.26 -22.9 > " > > Rémi > > _______________________________________________ > computer-go mailing list > [email protected] > http://www.computer-go.org/mailman/listinfo/computer-go/ > _______________________________________________ computer-go mailing list [email protected] http://www.computer-go.org/mailman/listinfo/computer-go/
