Hello Jeff, thank you for giving me the opportunity to elaborate on this.
Ingo: > >> Exhibition games from the years 2004-2007, during the Mainz chess > >> festivals, indicated that human grandmasters seemed to do easier > >> against bots in Chess960 compared with normal chess. Jeff: > I wouldn't take it seriously without further data. Going by the > Wikipedia page: http://en.wikipedia.org/wiki/Chess960 > > 2004: "At the same tournament in 2004, Aronian played two Chess960 > games against the Dutch computer chess program The Baron, developed > by Richard Pijl. Both games ended in a draw. Right. Observe that these games (all in Mainz) were played at quick times: Each side had 25 minutes in total for all moves. Fast play favours bots in chess. I was referee in the Mainz tournaments in 2005 and 2006. > It was the first ever man against machine match in Chess960. Yes. But already in 1997 I had played an 8 games match (with slow times; 3 minutes per move in the average) against Germany's no.1 GM, Artur Yusupov, in shuffle chess without Fischer castling. In 5 of the 8 games Yusupov had an advantage after the opening, and in two others the position was balanced. Yusupov had a score of +1, =4, -3 against my "List-3-Hirn". Two of his 3 losses were in the last two rounds, when he was tired (we played 8 days, one game per day). You can find details in my book "13 Jahre 3-Hirn" (in German; title to English: 13 years of 3-Hirn"). http://www.3-hirn-verlag.de/books.html From the Yusupov match I have appended a short episode on a pre-test, wehre we only played 12 moves - and Yusupov had a clearly won position after that. > Zoltán Almási won the Chess960 open tournament in 2004." In "normal" chess his r ating was around 2600 (at tournament time levels). > Problem: "The Baron" wasn't a top chess program. An archive page of the > computer chess rating list from 2006, the earliest I could find, shows > it at around 2500, > whereas top program, Rybka, from the same list is around 3000. Rybka (now declared to have been a semi-clone of Fruit, by the way) entered the scene only in December 2005. The Mainz tournament was in July 2005. > 2005: "In 2005, The Baron played two Chess960 games against Chess960 > World Champion Peter Svidler; Svidler won 1.5–0.5. The chess program > Shredder, developed by Stefan Meyer-Kahlen from Düsseldorf, Germany, > played two games against Zoltán Almási from Hungary; Shredder won > 2–0." > Again, "The Baron" lost, but Shredder (rated around 2850 from the same > list) won both its games. Unfortunately I do not have the Almasi games at hand. But I remember well that Almasi had a very positive position in the first game, before he blundered in time trouble. In the second game he was "denerved". > 2006-2007: Games between computers and grandmasters aren't listed. I'm > suspicious that this isn't a coincidence once the top programs got into > the field. It had to do with the overall schedule of the Mainz festival. Mainz was an event with almost 1,000 humans playing, and the organizers had other priorities. > If you've got other evidence, Ingo, please share. > Given that humans rely so much on experience, and computers are so > good at tactics, it would be surprising if Chess960 was more difficult > for computers. Chess computers have big opening books. Without them they sometimes play really very strange moves in the very first positions - being unable to repair the damage later. > I'd also think > computers would have a big advantage in creating an opening database > for all 960 positions if the programmers so chose to create one. Programmers had not done that in 2004-2006. (I know from my job and my discussions during those tournaments.) In the appendix you also find two photos from Mainz: one showing (from left) GM Svidler - Ingo Althofer - programmer the second one showing organizer Mark Vogelgesang and programmer Richard Pijl. Ingo. -- NEU: FreePhone - kostenlos mobil telefonieren! Jetzt informieren: http://www.gmx.net/de/go/freephone _______________________________________________ Computer-go mailing list [email protected] http://dvandva.org/cgi-bin/mailman/listinfo/computer-go
