Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Unmasking the Secrets of Rybka and Fruit

Author: Dann Corbit

Date: 13:21:15 12/15/05

Go up one level in this thread


On December 15, 2005 at 16:07:10, Stuart Cracraft wrote:

>Recently two programs came upon the scene and astonished many
>with their great results.
>
>Why do you think they do better, specifically?

Everything.

They evaluate better.
They search better.

Interestingly, I think that Rybka is an affirmation of Chris W. and Vincent D.'s
approach (pack as much knowledge as possible into the evaluation).  I say that
because it solves problems in much earlier plies than other programs do.  I
suppose that this could also be partly due to a better search.

Because Rybka does stupendously better in a 64 bit compile, I surmise that Rybka
is a bitboard program.  This is surprising because most of the top commercial
programs and Fruit are not bitboard programs.

Fruit is open source.  So we can figure out what Fabien does that is more clever
than our attempts (eventually).

I think Rybka will tip the see-saw again.

When Christophe Theron got great success with search speed, many people decided
that eval was not that important and concentrated on search.  Also, many highly
successful chess programs were not bitboard based, and so people started writing
0x88 and other matrix based chess programs.

Now, they will probably try to pack in more chess knowledge and go bitboard.

Of course, it's not going to help as much as most will hope.  The real thing
that makes the programs better is superior thinking and implementations of the
ideas of the authors.  The nuts and bolts of how to get there are not all that
important (IMO-YMMV).



This page took 0.01 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.