Author: José de Jesús García Ruvalcaba
Date: 10:40:49 06/01/02
Go up one level in this thread
On May 31, 2002 at 11:59:16, Chris Carson wrote: >On May 31, 2002 at 10:48:02, José de Jesús García Ruvalcaba wrote: > >> >>The following two facts have definitely caused rating inflation: >>1. For a long period of time the winner of a tournament did not lose rating >>points, even if she/he scored under the expectations. I do not know how many >>points were globally added to the list, but I think they were a lot. > >When was this. Was this in the beginning of ratings? If this was in the >beginning, then I would think ratings could have deflated due to this. It may >not be significant in either case. > It was not from the beginning, it took some years to introduce this regulation; and it remained a long time. In any case it can not cause any deflation, only inflation. >>2. FIDE gave once a gift of 100 rating points to almost all the women in the >>rating list (I think the only exceptions were Judith and Zsuzsa Polgar). Here I >>am sure there were a lot rating points in total. > >This could be significant. It depends on how many games these people played Yes. Players who quitted playing after receiving those points have no effect at all. > and >if they lost the 100 points/player. No, definitely not. It affects even if the women who received the points kept them after playing some amount of games. > This could also be measured, but I do not >think it is significant. > > >> >>The following factors can produce rating inflation or deflation, but I believe >>it is mostly inflation: >>1. The k factor is 15 for players rated up to 2400 and 10 for higher rated >>players. This can cause global lose or gain of rating points in tournaments >>where there are both players above and below 2400. I think it is mostly gain due >>to the abundance of young, talented players which improve very quickly and which >>are underrated for some tournaments. > >Has the k factor changed over time. If not, then I do not think it would cause >inflation. Could be significant, but I would need to see a proof. > First it was 10 for all the players, then it was increased to 15 for players rated less than 2400. Having different k factors for different players can definitely cause inflation and deflation. The simplest case is a match between two players, A and B. Assume A has a rating less than 2400 (and thus his k factor is 15) and B is over 2400 (his k factor is 10). If A scores above her/his expectation then there is some inflation, if B scores above her/his expectation then there is deflation. In tournaments (with more players) it is somewhat more complex, but the principle holds. > >>2. Players who drop the list (they become inactive or die) take their rating >>points with them. They could have won or lost rating points since they became >>their first rating. Let's call a "succesful player" somebody who wins rating >>points after becoming a rating, and a "unsuccesful player" somebody who loses >>rating points. Unsuccesful players are much more likely to quickly become >>inactive, in the meantime only (or mostly) giving rating points up, they cause >>inflation. Of course succesful players also become eventually inactive causing >>deflation, but since they tend to remain a lot longer (and they are way less >>than the unsuccesful ones), the global effect is smaller. >>José. > >Players dropping off the list or added to the list is not significant to >increase ratings. I would need proof to believe this. > Sorry, no proof for that, I was guided by my intuition. There is a good chance you are right. >All the above can be tested, but I do not think any have significantly inflated >ratings. The first three points can definitely be tested, it is only neccesary to recalculate all the rated tournaments with the original rules, and then make a comparison with the current rating list. I do not know if FIDE keeps a complete register over all these years, but if they are saved on a computer in a convenient format it would be a matter of a few hours on a current machine. The fourth point, about players who drop off, is not easy to test. We would need to know the results they would have got had they kept playing, which is open to wild speculations. > I do think that Fisher would have increased his rating if he had >stayed active. I think he would have been in the 2850 range, but not because of >inflated ratings, just because he was that good. He retired and we will never >know, however, ratings can be compared, that has a sound math/stat proof, see >ELO research. I can trust Prof. Elo, but I can not trust all the manipulations FIDE has done with the ratings. My criticism was not directed against the rating system devised by Elo, but against how it has been handled and mishandled over the years. Apart from those strange rules (whose introduction had to do more with politics), FIDE has manipulated the ratings more than once (and the award of titles too, but that is another topic). José.
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.