Computer Chess Club Archives


Search

Terms

Messages

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

Author: Bruce Moreland

Date: 15:23:59 05/25/00

Go up one level in this thread


On May 25, 2000 at 18:19:57, Tom Kerrigan wrote:

>On May 25, 2000 at 15:55:03, Ernst A. Heinz wrote:
>
>>>Sounds to me like you should test this for the same reason
>>>that you did the "X <=> X+1" test.
>>
>>Sounds to me that you still do not accept or understand the
>>qualitative difference between the cases ...
>>
>>=Ernst=
>
>I'm not even suggesting that the cases are remotely similar.
>
>I'm simply continuing the point of this thread, namely, you noticed that as
>depth increases, the number of draws increases and the number of wins decreases.
>Your conclusion is that this behavior is due to unequal depths. But wouldn't it
>be interesting if the behavior also occurred with equal depths?
>
>Here's the problem in my mind:
>You didn't believe that "X <=> X+1" would vary with depth, so you did an
>experiment and wrote a paper.
>You don't believe that "X <=> X" varies with depth, so you are dimissing the
>possibility with curt remarks and implications that I don't understand the
>problem.
>
>-Tom

I haven't fought with Ernst's gzip + postscript yet, but from reading other
posts in this thread it sounds like he did an 8 vs 8.  I don't think it is
necessary for him to do a 9 vs 9, etc.  That'd be like giving your control group
different doses of placebo.  You shouldn't have to do that.

bruce





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.