Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: informant symbol values

Author: Don Dailey

Date: 09:22:04 05/03/98

Go up one level in this thread


On May 03, 1998 at 03:17:45, Andrew Templeton wrote:

>   I would like to write a program that cleans up the analysis added to
>a PGN game by Rebel9 to make it easier to read. Assuming that most chess
>programs use 1.00 as the value of a pawn (at least I believe this is
>true) what do you think should be the value for the following chess
>symbols:
>
>equality  (=)
>0.00
>
>slight advantage (+/= or =/+)
>?
>
>moderate advantage (+/- or -/+)
>?
>
>decisive advantage (+- or -+)
>?
>
>   I would like to use the above symbols in the analysis where
>appropriate. I don't intend to replace the numeric evaluation, just
>enhance it. I could of course specify a range in an INI file but I was
>wondering if there was already any generally agreed upon values. If so,
>I would naturally prefer to stick with them.
>   Any thoughts on these, or the use of other symbols, would be
>gratefully appreciated.
>
>   Thanks,
>
>   Andy


You'll probably get some good feedback here.  I'm not aware of any
standard accepted values and they would vary from program to program.

But here is an idea that may help.  Calibrate these glyphs against
known human analysis to find out what reasonable thresholds are.

Take a sample of a few opening with master evaluation.  Look for a
few of each ( = += and +- )

Run your favorite programs to some reasonable depth and observe
where the transition is (compared to human assessments.) There will
probably be a lot of room for error so you may need to run quite a few
positions.   Then you might note the scores that seem to be on
borderlines of categories or perhaps overlapping and use this to
figure out a reasonable value.

Probably you should use this process on more than 1 program to get
a feel for the likely error.  It may turn out that the numbers that
make great sense for one program do not for another.   But if you
are only interested in Rebel 9 analysis there should be no problem.

- Don



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.