Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Junior - Crafty NPS Challenge - a user experiment

Author: Dieter Buerssner

Date: 13:00:44 11/22/03

Go up one level in this thread


On November 22, 2003 at 15:48:33, Robert Hyatt wrote:

>On November 22, 2003 at 03:52:25, Gian-Carlo Pascutto wrote:
>
>>On November 22, 2003 at 00:55:50, Slater Wold wrote:
>>
>>>If the 233mhz doesn't die first, here's my prediction:
>>>
>>>You play MORE than 12 games. Crafty wins, but with so many draws, >it'll "cancel" any *real* proof the match might have given.
>>
>>Nonsense. If he plays on till the score DIFFERENCE is 6 points,
>>it's staticstically a very valid result (somewhere >90 or >95%
>>certainty).
>
>The point is that if Crafty wins 6 games after playing 6 games, you
>get _one_ rating.  If it wins 6 games out of 1000 games, you get
>a totally different rating that doesn't prove much about the +6 game
>edge...

For the rating - yes. For the question "how certain can I be that the winner is
really better", it depends. When the remaining 994 games were draw, the
confidince level for Crafty beeing better (in the sense: will win any match) is
just as good as in a plain 6-0 without any draws. The rating will not show that.
The example might be not so good, because one really cannot calculate a rating
difference with 6-0. But say 11-1 and 1011-1001 give the same confidence of the
winner beeing better when in the first case it was +11 =0 -1 and in the second
case +11 =2000 -1. The calculated rating difference (which tries to predict
scores of matches, not the probability that one is better) will be very
different again.

Regards,
Dieter



This page took 0.01 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.