Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Performance rating calculation

Author: Dan Wulff

Date: 09:39:57 06/03/04

Go up one level in this thread


Hi Mike!

On June 02, 2004 at 20:34:36, Mike Byrne wrote:

>>there are other ways to do this, but this is very simple to set up in a
>>speadsheet as you mulitly the bonus by the result (1,0,1/2) add the base (opp
>>rating -400) to get the individual game performace rating and then just average
>>the performace ratings

This method would explain why one game can give a huge performance rating
difference. I'm not sure that this is a feasible method, though.

What I tried was the following:

A 2500 player (white) against a 2600 player (black):

The expected score is calculated as:

1/(1+10^((2600-2500)/400) = 0.3599 for the stronger player.

if the weaker player wins, he gets 16*(1-0.3599) = 10 ELO (Rounded down from
10.25, so his performance rating is 2500+10 = 2510 in the game and the stronger
player's performance is  2600-10 = 2590, which is clearly not what intended.
This is the way it is done in the ELO system, if I am not totally mistaken. A
draw would yield a difference of 2.24 rating points using this formula, but that
result is clearly misleading.

I'm confused.

Greetings

Dan Wulff (The Gandalf Team)



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.