Newsgroups: rec.games.chess.misc Path: info!dregis From: dregis@exeter.ac.uk (D.Regis) Subject: Re: FIDE ELO FORMULA Message-ID: Organization: University of Exeter, UK. References: <01bc1fd8$faf839e0$af4e8280@pc29.geoinfo.tuwien.ac.at> Date: Thu, 27 Feb 1997 14:00:02 GMT In article <01bc1fd8$faf839e0$af4e8280@pc29.geoinfo.tuwien.ac.at> "Damir Medak" writes: > >Hi, >I am looking for exact mathematical formulae of FIDE ELO calculation. >Can anybody help me? >Thanks a lot! >Damir There is a helpful description of the linear approximation formaula in a text file by Joe Brooks which is kicking around the 'net: "For those who are mathematically inclined and interested, the formula for a provisional chess rating is: 400 (W-L) Rp = Rc + --------- N where: Rp = the player's provisional or performance rating Rc = the average rating of the player's opponents W = the number of wins L = the number of losses N = the total number of games played (N <= 20 for a provisional rating) A draw counts as half a win and half a loss." Joe's text doesn't cover whatever the PCA do/did for their ratings. The formula comes from "normal" statistics which Prof. Elo used, setting the mean to 1500 and the class interval at 200. The formula above describes the expected result for a given excess of your rating over opponent(s)'s rating(s) and is an approximation to the normal distribution function, which looks a bit like this (on my screen at least!): ******* 1.0 ***** **** *** ** ** ** | 0.5 ** | ** | *** | **** | ***** | ***** | 0.0 roughly straight part of the curve where ratings are same or near each other That will be tuppence -- May your pieces harmonise with your Pawn structure and your sacrifices be sound in all variations D _ / "()/~ Dave Regis &8^D* Exeter Chess Coaching Page etc.: || \_/| = DrDave on BICS http://www.ex.ac.uk/~dregis/DR/chess.html ~\ / "...what else exists in the world but chess?" _|||__SHEU ~/sheu.html -- NABOKOV "Contribute!" -- Doug Attig From info!dregis Thu Feb 27 16:40:26 GMT 1997 Article: 16982 of rec.games.chess.misc