Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Why "positional" test positions are physically impossible! WM Test

Author: Robert Hyatt

Date: 12:54:23 12/03/02

Go up one level in this thread


On December 03, 2002 at 13:00:52, Thorsten Czub wrote:

>positional test suites are NOT impossible.
>
>the michael reiter test suites measured that.


Not very well.  A program that only understands one particular positional idea
can still
solve other types of positions for the wrong reason.  And inflate their test
results...

When they really know nothing at all.

For tactical tests, you can prove that you see the correct result, because it is
a search space
problem with a finite and measurable solution.

For positional tests, you can prove you don't misunderstand things by not
playing bad moves
in specific positions, but that doesn't prove that you _do_ understand what is
going on...



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.