Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Small number statistics and small differences

Author: Peter Fendrich

Date: 09:03:30 08/15/98

Go up one level in this thread


On August 15, 1998 at 03:43:02, Dan Homan wrote:

>Bruce,
>
>For 2 programs with a draw chance D
>  Pa = chance of program A winning any one game
>  Pb = 1-D-Pa =  chance of program B winning any one game
>
>For a 4-0 result in a 4 game match, either A or B has to
>win all games, so the probability, P, that there is a 4-0
>result is...
>
>P = Pa^4 + Pb^4
>
>So if the programs are equal, Pa = 0.4 and you get a 4-0 (or 0-4)
>result 5% of the time.
>
>If there is a 4-0 result, what are the chances that the stronger
>program won?  Assume A is stronger. The chance that A wins any
>4-0 (or 0-4) result is simply Pa^4/(Pa^4+Pb^4).
>
>So 4-0 results happen about 5% of the time between nearly

That's true for sure but you still don't know the confidence of the result.
Results of the type N-0 are very special because the guy with 0 is in the state
of 'free falling' without knowing where the ground is. The 4-0 result could be
the 4 first games of a 100-0 result. It could as well be the first 4 games of a
75-25 result. The difference between these two cases is huge because the 100-0
result is still a 'free falling' result. It could as well mean 1000-0...
In the 75-25 result we have at least seen the ground...

//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.