Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: WHAT IS THIS?

Author: Randall Shane

Date: 09:36:09 10/21/05

Go up one level in this thread


On October 21, 2005 at 11:36:38, chandler yergin wrote:

>
>There are no Bugs in any Engine!
>The Engine does nothing.
>The Algorithm does everything!
>
>http://www.seanet.com/~brucemo/topics/minmax.htm
>
<Bruce Moreland's text elided for conciseness's sake>

>
>To say there is a Bug in the Engine is just not correct!
>To say that the Algorithm is biased.. does not work for some positions
>is not correct.
>If it did, it would not work for any position, and would be disfunctional.

Hmm.  Might as well take a shot at this.

Chandler, there are numerous errors in your statement above.

1) Bruce's text was a basic outline of a class of algorithms based
around the minimax principle, not an algorithm itself.  It was
intended as a basic description for the beginner.  It certainly wasn't
meant as a detailed description of the internals of a wrold-class
chess program.

2) When you say that "The Engine does nothing. The Algorithm does
everything!", do you have any idea what that means?

An algorithm is merely a description of a process, of how to do
something.  A chess engine is a particular implementation in CODE of
an chess search algorithm In the computer world, algorithms --
especially loosely-described ones -- don't DO anything.  CODE does
stuff.

The basic alphabeta algorithm is modified in most engines with
quiescence searches, pruning, extensions, move ordering -- the
possibilities are seemingly endless.  The are a myriad of choices to
make in the implementation of the search.  Also, alpha-beta depends on
an evaluation function at each node of the search tree -- and there
are millions of choices to make in the design and implementation of
the evaluation function.  Some of those choices are demonstrably bad
Furthermore, implementations of algorthms in CODE can have bugs -- and
many of those bugs can be subtle.

Bugs can be obvious like program crashes, or less obvious such as
position misevaluation.  Fritz appears to have a bug in some
positions.  Chessbase agrees, and reportedly has issued a patch.

So, when you said that Fritz didn't have a bug, despite the evidence
-- well, you were wrong.  And using Shredder to prove your point,
well, doesn't prove your point, and was irrelevant.

If what you said was correct (and it isn't!) -- then wouldn't the top
chess programs play identically?

Fritz and Shredder are two different programs, written by different
people, with different design decisions, different search choices,
different evaluation functions, etc.  It would be beyond shocking if
they acted the same in all situations.  In fact, if two programs in a
complicated area such as chess produce identical output, that would be
serious evidence that some sort of cheating (such a stealing code) was
going on.

As always, the devil is in the details, Chandler.





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.