Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Rybka's current exe size: 4 628 480 !

Author: enrico carrisco

Date: 04:12:15 01/29/06

Go up one level in this thread


On January 29, 2006 at 06:46:35, Ryan B. wrote:

>On January 29, 2006 at 06:23:45, Jouni Uski wrote:
>
>>It's stunnigly big if we compare to other top engines:
>>
>>Toga     163 840 !
>>Fritz    446 464
>>Fruit    704 512
>>Shredder 802 816
>>
>>Can any programmer give a quess, what's the reason? Is it simply a lot of
>>knowledge? Or bad compiler? Totally different approach?
>>
>>Jouni
>
>The size is mostly a bunch of tables.  Disassembled the code is about 6.4MB and
>decompiled and cleaned up a little it is still about 3MB.  It is defiantly not
>chess knowledge.  The eval is actually very fruit like.
>
>Ryan

Hello Ryan.

Did you notice any "tactical-theme" based hidden assymetrical search?  It seems
a "fruit-like" 10-ply primary search could work with an added N ply of a very
fast search with captures and checks, etc.

That would explain its "long think" on positions when there are lots of
captures.  The very fast final search does not do a full eval but only does some
table look-ups based on pre-initialized data at 10-ply, for example.

The full PV or the garbage (positionally) at the end of the line can't be shown
or the cat's out of the bag.  This extra layer of search, however, does give it
a very good tactical ability.

Reminds me of Deep Thought -- using the hardware for the last N plies.  This
type of tactical search works real efficiently to see danger from your opponent
but less efficient in finding chances for itself (ex: Genius.)  Tactically it
makes it very strong but not so efficient in king attacks compared to Fritz or
Hiarcs.  Hence, on test positions it does slightly worse (just like Fruit.)

Regards,

-elc.









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.