Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: The Limits of Positional Knowledge

Author: Bruce Moreland

Date: 01:24:25 11/26/99

Go up one level in this thread


On November 24, 1999 at 00:36:41, Michael Neish wrote:

>Having recently written my first program, which can search only up to 6-ply,
>that's if you want it to move sometime before the big crunch, I thought what
>would happen if I stubbornly kept to a 6-ply limit and just tried to improve
>the program's playing strength by giving it increasingly better positional
>sense?

I'm going to speculate a lot, assuming the normal chess program model involving
an eval function at the tips, rather than an eval function that guides the
search.

I think that in this case eval is like the tread pattern on tires, and search is
like the drive train.  If there's no tread on the tires, you can't apply power
to the road.  If the tread pattern is seriously wrong, you'll have a hard time
controlling the car and you might go off the road.  If the tread pattern is
right, the power goes to the road and you are happy.  If the tread pattern is
super-wonderful, then perhaps things work a little better, but you'd get more
advantage if you improved the drive train.

So I think that if you stick to six plies you'll get your butt kicked by
programs that search much deeper than this.  You'll lose on tactics in positions
that are better, and you'll be forced to make positional concessions in order to
avoid tactical consequences that you barely see, and you'll lose because of
this.

There are positions that you might understand perfectly but in most cases you'll
still lose because of the incredible tenacity of the deep searchers.  It is hard
for a creative guesser to defeat a diligent researcher consistently, and that is
what you'd be trying to do.

I think you'd do a little better against humans than you'd expect from your
rating against computers though.  Fatal tactical mistakes against computers
often turn out to be excellent positional bluffs against humans.

bruce




This page took 0.01 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.