Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Two interesting positions to test your Programs Materialism

Author: Bob Durrett

Date: 16:22:23 11/12/02

Go up one level in this thread


On November 12, 2002 at 18:20:54, Uri Blass wrote:

>On November 12, 2002 at 17:56:14, Bob Durrett wrote:
>
>>On November 12, 2002 at 17:49:48, Uri Blass wrote:
>>
>>>On November 12, 2002 at 17:39:55, Bob Durrett wrote:
>>>
>>>>
>>>>
>>>>Without using a chess engine to check my intuition, my first impression is that
>>>>king safety issues dominate.  If true, then this is really a test of the
>>>>engine's ability to recognize that fact.
>>>>
>>>>Bob D.
>>>
>>>No
>>>
>>>Movei has no king safety evaluation
>>
>>?????????????????
>>
>>My chin fell to the floor when I read that!  You do not have ANYTHING in your
>>program at all having to do with king safety???
>>
>>Again relying on my "naked intuition," I would have thought that no chess engine
>>could even exist without that.
>>
>>King safety is "lesson number one" in teaching of HUMAN chess players.
>>
>>Bob D.
>
>It knows that it should not go forward with the king except endgame thanks to
>piece square table but not more than it.
>
>Note that I do not try to teach humans to play chess but to teach my program to
>play chess better.
>
>I plan to teach it about king safety in the future but today I improve it by
>other ways.

That makes a lot of sense.  Good idea. : )

Nevertheless, I still find it remarkable that your program can do well without
SOMETHING at least distantly related to king safety.  Maybe the answer is:  "If
you play good chess, you won't get checkmated."

Bob D.


>
>I do not like to teach it simple ideas like no pawns near the king are bad(a lot
>of amateurs use this rule for their program) because there are cases when it is
>not bad.
>
>Maybe I can get some improvement by this rule but I do not like writing
>evaluation code only to delete it later.
>
>inspite of lack of knowledge Movei has chances against everything and recently
>it drew a blitz game against ruffian(the best freeware program) and it was
>not a grandmaster draw.
>
>see http://www.geocities.com/cpf27/page9.htm for the tournament and you can
>download the games(The movei that was used there is not the latest movei but it
>is clearly better than the public version)
>
>Uri



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.