Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: To ALL: Look at Christophe Answer to my post about Gambit jumps...(NT)

Author: Robert Hyatt

Date: 13:26:50 04/04/01

Go up one level in this thread


On April 04, 2001 at 15:42:43, Ulrich Tuerke wrote:

>This "2nd order evaluation" stuff is not as sensational as you may think,
>Fernando.
>Any program, using more than just piece square tables, has 2nd order terms. For
>instance giving a bonus for a knight occupying an opponent's weak square is a
>2nd order term, because  the weak square is a consequnce of the pawn structure.
>Is there a program which does not account for a knight forepost ?
>
>IMHO, these large fluctuations in score which you have observed, are just a
>consequence of using rather large values in the static evaluation function. Any
>program giving large values for king safety, passed pawns, ... shows these
>effects. Even mine, but I'm not always completely happy with it.
>
>After all, I'm also curious to test the Gambit Tiger because I find it always
>intersting to play these kind of programs with kind of speculative evaluations.
>And it seems that Chris had a very lucky and skilful hand for this, because
>Tiger is successful in spite of this speculative eval.
>
>Just my opinion, may of course be completely wrong.
>
>Regards, Uli


I am not sure why that raised any eyebrows at all.  2nd order evaluation terms
have been around for way over 30 years...  since the first decent chess
programs.

I go one level further in fact.  IE for king safety, I first analyze pawn safety
around the castled king, then I use this to analyze pieces around the king, and
finally a third-order evaluation that folds in all 3 together so that more
pieces around a king with weak pawns is considered good...

And the entire mess is sent thru an exponential function so that going from two
to three pieces around the king adds more than going from 0 to 1 or 1 to 2.
And more pieces around a wrecked king position is better than even more pieces
around a very safe king position...

This seems like a natural way to evaluate a position...  And can let you tune
things very aggressively if you want to.



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.