Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: bean counters

Author: Peter Berger

Date: 12:38:20 11/16/00

Go up one level in this thread


On November 16, 2000 at 10:29:02, Robert Hyatt wrote:

>On November 16, 2000 at 09:14:38, Peter Berger wrote:
>
>>On November 16, 2000 at 09:07:23, walter irvin wrote:
>>
>>>to me programs fall into 2 list bean counters and knowledge based .
>>>bean counters
>>>fritz
>>>junior
>>>nimzo
>>>lg2000a
>>>
>>>knowledge based
>>>shredder
>>>hiarcs
>>>rebel
>>>tiger
>>>diep
>>>crafty
>>>king
>>>
>><snip>
>>
>>Could you post an example for a position where for example Crafty shows superior
>>chessknowledge compaired to Junior ?
>>
>>pete
>
>
>Since I don't have any commercial program, I can't test them, but here is one
>simple test, just for fun.  Evaluation after 1 ply search...
>
>[D]4k///3pp/1P4P///4K w
>
>This is just one test, of many.  The idea is that last year, a GM friend of
>mine tried this against a couple of well-known programs, and both preferred
>black with the two connected passed pawns.  This would be correct if there
>were lots of pieces left on the board.  But in this position, the white king
>can stop the black pawns easily.  The black king has a problem.
>
>It would be interesting to see how other programs handle this.  My static
>eval is +1.5, for instance, although a 1 ply search will produce the same
>score, approximately.
>
>If you want, we can try a few of these for fun, for anyone willing to
>participate.  I hate to give away all my 'secrets' of course, but since Crafty
>is open source, nothing is really secret. :)

I think this discussion is definitely very interesting for me probably because
the topic already shifted : we are now not talking about "knowledge based" but
about being "knowledgeable" ( not sure about this word : "having knowledge" is
what I mean ) now .

But first a little step aside :

My theorie was that if one program is "knowledge based" and the other is "bean
counter" there should at least be one position where the "knowledge based" one
plays a move showing super chessknowledge . As it seemed to me putting the
programs in these two sections is a pretty random job I assumed it might be hard
to find one ( especially with the two programs chosen ) . I didn't think
thoroughly enough before posting this one-liner though as even if there would be
_no_ position where the "knowledge based" one shows superior chessknowledge on
the board it would still say nothing about its being "knowledge based" or not .
And please don't get me wrong : I am not trying to say anything bad about crafty
here ; remember my post last week about a position Crafty understood clearly
better than Gambit Tiger.

"Bean counter" seems to be a more or less understandable term . In my opinion
all current programs seem to fall into this category .

"Knowledge based" is undefined in chess context as far as I know and probably
doesn't make too much sense anyway . I read an article about a very early
chessprogram that tried to find the best move by asking and answering questions
to itself : Bernstein , and about one that formulated aims before starting to
search : NSS , but I am unsure if this has anything to do with being "Knowledge
based"  or with the article which was written in popular style.

I personally prefer the A-M und N-Z approach :) .

An idea : maybe a "knowledge based" program should simulate a thought process
like the one descibed by Kotov "Think like a Grandmaster" or Silman "How to
reassess your Chess" . Especially with Silman it sometimes sounds codeable but
his trick is that he leaves certain terms he uses all the time undefined ( as
they are understandable for human intuition ) which might make it hard work :) .

There _are_ programming ideas like Pattern Learning or Self-teaching or whatever
which are sometimes mentioned here but as far as I know haven't produced a
competitive chessprogram so far .

OK , but I am in deep water here talking about things I barely understand anyway
.

Most of the time "knowledge based" is used in this forum for programs with low
NPS and displaying high evaluations . A good test would be if one programmer of
a well-known "bean counter" released a new version _displaying_ NPS/4 and EVAL*3
as long as material is even . I wonder if people would suddenly think its
knowledge has improved and call it "knowledge based" . Also if a programmer
found a way to only search a tree half as big with NPS going down by a 1/3 which
sounds like a good idea : is this "bean counting" or is this "knowledge based
idea" .

Back to the initial topic. A question which is valid to me though and might be
of use to both programmers and users is which program plays most knowledgeable
moves and makes best decision based on them . A good sideeffect is that others
might decide to code the same knowledge in later :)

I use my chessprograms mainly for long analysis and want to learn to improve my
judgment ; therefor eval of positions is important for me ; time until move
found usually of less importance .

It is fun to compair judgments of top programs : everyone seems to have a field
he rules in but on the other hand regions where the misjudgement is still
obvious enough to be understood by a mediocre player ( like an Expert player in
US terms ) .

How a program reaches its good decisions is of less importance IMHO . If the
moves are the best I don't care why it chose it .

There are several positions where knowledge is important and a program not
having it can _never_ find the right moves ( several examples posted here ) . I
recently tried a pawn endgame ( White win ) which was thorougly discussed on a
fast server with crafty letting it run for about 100000 minutes until ply 31
finished . It always had 0.00 as it was all about zugzwang .

The _displayed_ knowledge or approach is also important for the somehow silly
hobby of letting two programs play each other . If the two programs play
meaningful and interesting games where you can learn something it is fun . If
they only go around shuffling pieces not commiting any tactical mistakes until
one of it crashes it is boring and useless . So I currently like programs like
Gambit Tiger ( which is extremely well attacking ) or Shredder ( which is
extremely well defending ) . It opens new horizons.

I also think that Fritz6a sometimes plays very interesting and good games btw.

I will definitely have a look at your position and hope more of them are being
discussed . But for now I have to leave for my autumn holidays so my trial with
various programs will have to wait for 10 days :-)

Greetings.

pete












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.