Author: Stephen A. Boak
Date: 16:59:02 11/10/02
Go up one level in this thread
On November 10, 2002 at 08:27:55, José de Jesús García Ruvalcaba wrote: >On November 09, 2002 at 19:35:23, Stephen A. Boak wrote: > >>On November 09, 2002 at 13:59:43, Jorge Pichard wrote: >> >>This also hold true for lower rated players, if >>>you get a player with an Elo of 2200 and face him to play against a group with >>>an average Elo of 1800 it won't be unusual for him to get a performance rating >>>of 2400 or higher. >>> >>>Pichard. >> >>Traditional ELO performance rating calculations use the +/- 400 point rule. If >>you win all your games in the event, your *maximum* performance rating is >>limited (capped) at 400 points higher than the average of your opponents. Thus >>your scenario above would be impossible. >> >>I realize there are other software calculators that do not use the +/- 400 point >>rule, but then I don't consider them to be ELO based. [They may be considered >>alternative rating systems, and worthy of experimentation or implementation, but >>they are not really the ELO system.] >> >>--Steve > >Actually those 400 points do not come from Prof. Arpad Elo's original system. >They are used only for approximate (but fast) computations. >José. Jose, I loaned out my book by Arpad Elo a few years ago and never got it back. A pity, since I really liked that book. However, I recollect he addressed the topic--how to assign a provisional rating for starting players who have played only a few games. I think he suggested the +/- 400 rule in his book (if someone has the book, please correct me if I am wrong). Elo's view was that such a method, after enough provisional games, would approximate the true strength of the player, and then the players provisional ELO rating (using the +/- 400 rule) would become the permanent rating, subject then to other rules for changing permanenent ratings. In addition to the book by Arpad Elo, Elo instituted his formulas in the USCF (US Chess Federation) system back in the 1970's I believe, and the +/- 400 rule is and has been in effect there (for provisional players). I think Elo was the ratings director in charge of implementing his ELO system in the USCF. I don't remember exactly where the idea of using the +/- rule to determine a TPR (Tournament Performance Rating) surfaced first. I have, for years, used the +/- 400 rule to calculate a TPR figure--based on the provisional player concept that it approximates a player's true performance rating (given enough games) for the situation where there are a limited number of game results to be examined. I think it works well, even in 6 rounders (although all such TPR calculation methods are subject to debate on the pros & cons). Others have posted TPR figures using some sort of ELOstat or Chessbase ratings program, at times, and I noted many times that the TPR figures they posted didn't match those of the +/- 400 rule. Often their TPRs were much higher than that given by the +/- 400 rule. I never understood why those TPRs were used--they didn't look 'realistic'. If I enter a 6-round class tournament and win all my games (I've done that twice in my chess career), my performance rating is more likely the average of my opponents + 400 points (via the +/- 400 rule) than a far higher figure (let alone an infinitely positive figure!)--as indicated by the other types of TPR calculation results I've seen posted that would put my performance rating at 600-800+ points higher than my opponents. Elo believed that if you win or lose all your games, that your true strength is not known or knowable (I agree) and used the arbitrary 400 point difference as a limit. I believe it is not the only TPR calc method, but that it is more accurate than the other TPR results I have sometimes seen. For an established tournament player, say me for example, even when I have an exceptional result, that doesn't make it likely my performance was even close to that of a player 800 points higher in rating! The 400 point cap works very well in practice (although no TPR method is perfect, certainly). Thanks for the feedback, --Steve
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.