Author: Chris Carson
Date: 17:30:34 05/12/01
Go up one level in this thread
On May 12, 2001 at 19:06:00, Jeroen van Dorp wrote: >On May 12, 2001 at 10:48:20, Chris Carson wrote: > >>The SSDF is a valid and reliable measure of program strength on the >>hardware measured for tournament play against humans and other >>programs. >> >>The top programs have proven to perform above 2500 (2500 is the minimum >>for GM strength). Some of the programs have performances above 2600 >>and DJ6 has a 2702 performance against the top GM's (finished 50% >>score). > > >Hi Chris, > >Are you sure SSDF claims SSDF-ratings are exactly comparable with Elo-ratings? >As long as you have no humans in the rating pool, you cannot decide on the >starting (Elo)point for determining the rating - as you want to compare with >humans, after all. > >More decisive I think are indeed the results of chess programs participating in >tournaments and competition until now, scoring *very well* between solid IM's >*and* IGM's. > >Maybe - just maybe- if all chess players would become more accustomed to >computer participants in their tournaments, results could go down, because of >the diminishing "surprise" factor- so to say. > >BTW as human GM's are drawed by weaker players now and then, I think the same >goes for computer programs. I didn't mean to react on this discussion with the >gandalf game I posted somewhere else on this board - I hadn't read this thread - >but as you can see even a weakie like me has some luck, sometimes. > >Sometimes. Not often :)) > >There's much discussion about the desirability of computers in human >tournaments, and there are a lot of arguments pro and con, but I'm very >interested in regular participation to really validate the results until now. >I also think we might learn a lot from these games, "programming-wise". > >J. J., You bring up some good points and I am also very interested in continued human vs program/hw games at all levels. :) Yes, I think SSDF and ELO are fairly close (neither one is ever exact and both are in a state of flux). SSDF did calibration against humans (see the link posted) and has re-calibrated based on recent games (last spring if I recall). All the games are at 40/2 for the SSDF and the Human vs computer ratings. Just remember that a 400 point difference in rating implies that the weaker player should win 1 out of 10. I think many people forget that. That could also be 2 draws. If the result is different, then the ratings for both players change to incorporate the latest results (thus a state of ratings flux). Ratings change all the time, so expect some upsets and some rating changes. This is truly a great time to be interested in Human vs Computer Chess. In less than 5 years, it may be less competative. 5 years ago most people would laugh at the thought of a pc vs World Champ match and today the World Champ requires 3 months of preparation before he would play the program. Interesting that the program must provide the handicap. I am still amazed that Kramnik can not just show up and play his best game and be confident of a match win. :) Best Regards, Chris Carson
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.