Side effect of the countback system

General discussions about ratings.
Mike Gunn
Posts: 1014
Joined: Wed Apr 11, 2007 4:45 pm

Re: Side effect of the countback system

Post by Mike Gunn » Tue Jan 31, 2012 7:12 am

30 is based on statistical theory. A certain number of game results (n) will give you a x% probablility that your grade lies in the interval G-y to G+y. Elo has a table of y versus n (I think). I can't remember the details or the value of x Elo used or y when n=30. (I once tried working it out using my own very basic ststistics and couldn't get the same answer as Elo.)

Roger de Coverly
Posts: 21291
Joined: Tue Apr 15, 2008 2:51 pm

Re: Side effect of the countback system

Post by Roger de Coverly » Tue Jan 31, 2012 9:44 am

Mike Gunn wrote:30 is based on statistical theory. A certain number of game results (n) will give you a x% probablility that your grade lies in the interval G-y to G+y. Elo has a table of y versus n (I think). I can't remember the details or the value of x Elo used or y when n=30. (I once tried working it out using my own very basic ststistics and couldn't get the same answer as Elo.)
There's an independent derivation in this thread

http://www.ecforum.org.uk/viewtopic.php?f=4&t=361

The number 8 which appears there may be of relevance since the original design of the Clarke system published your grade as being between x and x+7.

Post Reply