Author: Uri Blass
Date: 13:01:51 08/26/00
Go up one level in this thread
On August 26, 2000 at 15:06:13, Stephen A. Boak wrote: >With an ELO-based system, after a multiple round tournament among >well-established players (i.e. rated over many games), it is possible to >calculate and say: > >Player A performed (scored) as well as a player of XXXX rating would have been >expected to score against the same opponents. > >Player B performed (scored) as well as a player of YYYY rating would have been >expected to score against the same opponents. > >One can then compare XXXX and YYYY to see what the results of that tournament >suggest about the relative ratings of Progam A and Program B. > >To apply this method to WMCC and get some meaningful conclusions about program >playing strengths is difficult at best (perhaps not impossible), since the >particular programs appearing at WMCC may not have well-established ratings due >to being recent software improvements over previously released and rated >versions. And those prior ratings of prior versions may have been earned on >different hardware platforms. > >Subject to some rating estimates (guesses) and shrewd thinking, it is possible >to assign some plausible ratings [but don't ask me to do this!] to the programs >that competed in the WMCC and then run the suggested ELO calculations. You can start with equal rating for all the players not inclusing players like Xinix and paque Expert(if you can divide the program to 2 teams when team A of program got 100% against team B then it is better not to include team B and in the case of the last WCCC there were 2 programs in team B) After it assume the result of the tournament happens again and again 1000000 times in a row and calculate the exact rating of all the players based on this assumption. Uri
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.