Author: martin fierz
Date: 13:58:03 04/26/02
Go up one level in this thread
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 ernst heinz had a formula for the error on a series of games in his book "scalable search in computer chess". i left it back in switzerland, so i can't look it up. it took into account win,loss and draw probability as taken from the match result and not only the overall win% - but i think there was also a simplified formula (which gave a larger error), which only used the win%. if somebody has this book at hand, he could look it up. of course, the error scales with 1/sqrt(N), as pointed out by ratko, but also of course, this is just the scaling, not the real value. there is something else in it. aloha martin
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.