Author: Mike Byrne
Date: 16:48:19 07/30/03
Go up one level in this thread
On July 30, 2003 at 06:08:08, Sherry Washington wrote: >Shredder is listed at 2800 on athlon 1.2, I don't think so. Personally I think >the list should be lowered by 75 points. Not sure if even the 8 processor fritz >is 2800 Not 2800 FIDE , but 2800 SSDF it is rated. There is a miconception that the SSDF reinforced many years ago. They had computers play some humans to "calibrate" the rating to the human scale. I'm not sure why, but I guess they felt the information would be more useful. Then they reinforced that notion again by lowering the ratings. Once again, I would presume that they felt a need to make the ratings look more human like. Many machines were downgrade by 100 points or so. Even if it did not play a game. The problem of course is that the machines are not playing humans. The machines are also getting faster every year or two. There is no way a program that is run on 1/4 the speed will keep up with a program running on a 4x machine - everything else being equal. With no human players in their little universe, the ratings are getting inflated rerlative to human players. One could argue that there is diminishing returns for faster hardware vs human players, but versus slower hardware - faster hardware is faster there is less noticable effect of diminshing returns. This phenomena has been also noted in very small pockets of human players that do not play anyone else from the outside. See related links about Claude Bloodgood, a master level player who became the second highest rated player in USCF while behind prison bars , playing just other prisoners. http://home.adelphia.net/~dwedding/chess/bloodgood/bloodgood.html and for those that want more detail http://www.asigc.it/teoria/Unorthodox_4.htm "Bloodgood continued to play offhand chess games within the prison walls throughout the 1980’s, and by the early 1990’s decided to re-form the VAPEN Chess club. This sparked yet another controversy that again drew national chess attention to Bloodgood. During this time, Bloodgood held tournaments inside of Powhatan Correctional Center. These tournaments uncovered a serious flaw in the United States Chess Federation (USCF) rating system. It turned out that the USCF provisional ratings formulas along with the ratings floors caused serious ratings inflation if a small group played numerous games within a closed pool. This is precisely what happened at the VAPEN chess club. As the VAPEN chess club was, what might euphemistically be called, a "captive audience" with not much else to do, Bloodgood and about twenty or so of his fellow inmates started playing several rated games a day. This group had members that were playing literally thousands of games a year. This caused their ratings to jump to extremely high levels. When it became obvious to Claude and VAPEN that the ratings formula was flawed, they alerted the USCF. Unfortunately, the USCF could not correct the problem quickly, and soon Claude Bloodgood was ranked second in the United States with a rating of 2702. This triggered an outcry from some members of the United States Chess Federation who alleged numerous unfair and untrue allegations that Bloodgood was committing ratings fraud. This was, of course, not the case as Bloodgood had alerted the USCF to the problem long before he ever reached the stratospheric level of 2702. Bloodgood was also quick to admit to anyone who would bother asking that he was over rated. His true ability at that time in his life was probably high expert class to low master level. At his prime, in the early 60’s, he was probably close to senior master strength." This is a very similiar situation that SSDF faces. Call it the "Bloodgood" effect. Small pool of players, thousnads of games, faster and faster and faster hardware. Rating inflation is "built-in" by unintentional design. But to reduce the ratings of the old established computers that actually played humans is not correct either. If they want comparision to humans, they must play humans. Otehrwise they should just leave it alone - the "Bloodgood" effect on computer ratings does not hurt anybody.
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.