Author: James T. Walker
Date: 13:09:25 10/13/99
Go up one level in this thread
On October 13, 1999 at 13:49:41, Ratko V Tomic wrote: >Thanks for the formula. Have some questions: > >> In my opinion the simplest formula is PR= Rc +400(W-L)/N. > >Is this valid only for relatively small difference in ratings (i.e. when W-L is >much smaller than N = W+L)? Looking at the other formula (with log) in this >thread, it seems this would work well only for "small" differences (maybe up to >2-3 hundred points). *********** As far as I know the formula works for all cases ************ > >When results against multiple players are used, is the PR averaged over all the >opponents (i.e. if there are C players, the right hand side is added for all of >them and divided by C)? Or is the average Rc computed and W's and L's added for >C opponents as single W and L totals and the formula applied only then? (The two >ways are not the same unless N is the same for all players.) PR is the Performance rating you are computing (There is only one PR). Rc is the average rating of the opponents. For example, 1800,1900,2000,2200,2100 = 100000/5 = 2000=Rc W is the number of wins by person/program you are computing PR for. (3) L is the number of losses by person/program you are computing PR for. (1) You can ignore draws since they cancel out but they must be counted in the total number of games played (N).(1) So for the above ratings: PR=2000 + (400(W-L)/N) = 2000 +(400(2)/5) = 2000 + 800/5 = 2000 + 160 = 2160. So if you play 5 people with an average rating of 2000 and you score 3 wins 1 loss and 1 draw, your PR would be 2160.
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.