Formula for calculating ELO performance

[Rein Halbersma]

See this paper for a clear explanation.

http://www.recherche.enac.fr/~alliot/elo.pdf

[Dann Corbit]

See this paper for a clear explanation.

http://www.recherche.enac.fr/~alliot/elo.pdf

Here is a very interesting article on alternative methods for Elo calculation by Roy Jones:
http://www.ratingtheory.com/

[Bill Rogers]

Well fellows, the math is a little beyond me, so who can say which system should be used or which system is the most accurate?
Just wondering.
Bill

[Dann Corbit]

Well fellows, the math is a little beyond me, so who can say which system should be used or which system is the most accurate?
Just wondering.
Bill

The answer to your question is very simple:

Which system makes the most accurate predictions? We can test this with experiments. The code from Royal Jones that I posted {prog10.cpp} does exactly that for Elo, linear and ratio models.

Besides contrived experiments, we can also model the predictions against actual data and see how often our predictions come true. Which ever method predicts most accurately is the best model.