Author: Sune Fischer
Date: 07:34:49 04/26/02
Go up one level in this thread
On April 26, 2002 at 10:23:32, Gian-Carlo Pascutto wrote: >On April 26, 2002 at 10:11:02, Jorge wrote: > >>On April 26, 2002 at 09:29:26, Gian-Carlo Pascutto wrote: >> >>>Given a score from a series of games, y, we can calculate >>>the ELO difference between the two players as: >>> >>>x = -400*log((1-y)/y) >>> >>>My question is now, how do we calculate the error margins >>>on this value? >>> >>>-- >>>GCP >> >>Maybe I am missing something here, but, this formula is not taking into account >>W,L,or D. So for example if total games were y=20, the ELO difference x is the >>same for a player that won 20-0 vs. a player that went 10-10? I don't get it. > >y = score, not number fo games > >There is a difference between scoring 20/0 (+inf) and 10/20 (=0.5) If I remember correctly then 0<=y<=1, so 20/0 would be a clean 1.0 (100%) score. of course the formula can't handle y=0 or y=1, but that formula is also only an approximation, the real formula can handle it. http://www.vogclub.com/ratings/elodetails.phtml When it seems to diverge, it is because a result of 20/0 translates into a 100 probability of a win for the guy who scored 20. Of cause such a probability is incorrect. -S. >-- >GCP
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.