Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Match results - a complete(!) theory (long)

Author: Peter Fendrich

Date: 07:18:30 02/19/01

Go up one level in this thread


On February 19, 2001 at 00:29:42, Ricardo Gibert wrote:

>On February 18, 2001 at 16:14:34, Peter Fendrich wrote:
>
>>
>>In another message Bruce Moreland raised the question
>>about how to interpret a match result.
>>With a match result between two players, he wanted to
>>express things like:
>>Player1 is better than Player2 with a certain significance
>>level.
>
>You can't do this with a *match* between the two players. It is possible that
>player A can consistently beat Player B in a match, but player B consistently
>has the higher rating as acquired against the general population of players.
>
>An example of how this would effect engine vs engine matches would be as
>follows: Program A has an opening book that is optimized to beat program B. Not
>surprisingly, program A edges out program B in any match played between the two,
>but if program B performs better than program A against other programs, then
>clearly program B is the "stronger" program even though program A is able to
>"dominate" B.
>
>When you measure the performance of player A against player B, you are measuring
>the ability of player A to "dominate" player B. You are *not* really measuring
>the playing strength of A. You do that by measuring its performance against a
>*population* of players.

It's true enough that it is the internal strength between
the two players that is meassured, but it's part of the
overall uncertainty in the long run. It's the best you can
do with just one match without ny prior knowledge about
Pw and Pd. If you have some information about how Pw and Pd
are distributed for the two players you could as well include
that. Patterns like the one you refer to are, as I mentioned
in the message, best captured by observation of the games and
we shouldn't rely too much in figures like this. It is however
interesting to get some indication of what the match result means
on condition that you have no other knowledge or if it's
included already.
//Peter



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.