Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A New Self-Play Experiment -- Diminishing Returns Shown with 95% Conf.

Author: Tom Kerrigan

Date: 10:53:40 05/26/00

Go up one level in this thread


On May 26, 2000 at 11:35:00, Ernst A. Heinz wrote:

>>I've heard many people complain that very strong human chess players draw each
>>other a lot. So it seems that the percentage of draws increases with human
>>strength. Is it that crazy to think that the percentage of draws also increases
>>with computer strength (i.e., depth)?
>
>No, it is not crazy at all -- my experiment actually shows this
>for "X+1 <=> X" slef-play of "Fritz 6" with 95% confidence.
>
>So, where is the beef?
>
>I was not interested in the question whether the number of draws
>also increases for "X <=> X". Why should I test it then as you
>repeatedly insisted?
>
>What influence does a higher number of draws for "9 <=> 9" than
>for "8 <=> 8" have on the question about the difference in
>playing strength between "8 <=> 7" and "9 <=> 8"?
>
>=Ernst=

Please do not be so loose with the language. I have not "repeatedly insisted"
anything. I haven't even insisted on something ONCE.

I see your point about the purpose of the experiment. (Too bad you didn't bring
this up 4 or 5 posts ago.)

Here's my explanation for the misunderstanding. You've run an experiment that's
a little weird, because one side searches a ply deeper than the other side. Then
you draw conclusions from the experimental data. It's easy to get confused and
think, "Ernst says that the percentage of draws increases _because_ one side is
searching slightly deeper than the other side." But if that's not what you're
trying to say, then I'm all good with it.

-Tom



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.