Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: The superior Rybka chess knowledge

Author: enrico carrisco

Date: 15:28:17 01/20/06

Go up one level in this thread


On January 20, 2006 at 16:14:31, Vincent Diepeveen wrote:

>On January 19, 2006 at 08:10:34, Vasik Rajlich wrote:
>
>>On January 18, 2006 at 13:34:17, Chrilly Donninger wrote:
>>
>>>After playing several engine matches against Rybka (chess programming is a
>>>rather boring job) I have come to the conclusion: There are a few special
>>>evaluation features of Rybka which are really unique. It is interesting that
>>>some seamingly relative unimportant feature appear regularily on the board. The
>>>opponent has no idea of this feature and does not prevent it. And the search
>>>always finds a way to reach the pattern. Rybka has e.g. some special passed pawn
>>>evaluation terms. I do not want to tell the details, but the game Zappa-Rybka,
>>>Paderborn 2005 is a prototype game for one of these special features.
>>>But the main chess-knowledge which sets Rybka appart from other engines is
>>>ignorance. The omission of features which other engines have incorporated.
>>>I have written a longer article for the German "Schachkalender 2006". The
>>>message of this article is: Most of the published chess knowledge is completly
>>>useless. Give your favorite chess-enemy your chess-books as a present. They will
>>>do some harm on his play.
>>>Rybka seems to be to prove of this hypothesis. If a feature is - in a given
>>>position - correct, it is of course an advantage if a programm has implemented
>>>it. But if its wrong, the programm hangs on an advantage which does no really
>>>not exist. Or even worse, it sacrificies another advantage to reach the pattern.
>>>
>>>I realized the principle: "It is sometimes more important to remove features
>>>than to add ones" several times in the Hydra project. E.g. Piece-Square Tables
>>>are generally considered as a "must have". Strong Chessplayers do not like them.
>>>It is very unnatural for them to evaluate a piece without considering the
>>>context of the other pieces. It took some time till GM Lutz convinced me to
>>>remove them in Hydra. And indead, the programm played considerably stronger with
>>>Piece-Square.
>>>Insofar is the Rybka approach intelligent ignorance.
>>>
>>>Chrilly
>>>
>>>P.S.: The omission of Piece-Square-Tables is a feature of Hydra. I do not state,
>>>that Rybka as skipped this feature too.
>>
>>Yes, what you omit is just as important of course.
>>
>>A game like Zappa-Rybka would be much more likely to be played as black by
>>someone like Fischer than someone like Tal. It's not because Tal doesn't know
>>the basics about a queenside pawn majority.
>
>Fischer played very classical openings, Rybka on other hand is good in modern
>positions.
>
>Fischer took big risks on kingside sometimes, Rybka never takes risks there.
>
>In that sense, Rybka is a very ugly player from human viewpoint are.
>
>Of course Rybka is not unique there. Most engines qualify there :)
>
>I remember Fernando comparing it with Karpov. However Karpov is a very
>positional player, and Rybka isn't. There is a dozen engines more positional as
>they have more knowledge (which of course means you could have potentially more
>bugs, i agree in that respect with Chrilly, but it also means they have more
>potential to play strong in future if those bugs ever get fixed). Shredder is
>100x closer to Karpov than Rybka is.
>
>Rybka compares well with the Richard Lang style programs from a decade ago, in
>term of style.
>
>What we do know already for many years is that tournaments in computerchess are
>won easier by passive bugfree play, than by agressive attacking play.
>
>Shredder and Genius are good examples there. Fritz3 in its glory days was very
>passive too.
>
>It is a race against bugs in computerchess always, in that sense you're doing a
>superb job currently with Rybka. Very bugfree play.
>
>>I suspect Hydra will have a harder time than Rybka playing a position like this,
>>but then again Rybka will have a harder time in some other types of positions.
>>Fortunately in chess we have a way to settle these things.
>>Vas
>
>Odds are of course zero that we'll ever meet Nimzo1998-on-stereoids or Fritz in
>an official tournament ever again. Well that is, unless someone manufacturers
>some real fast 'pocket pc' in which case Pocket-Fritz version 10.0 might win.
>But hey, we might also know that under a different name... ...as Shredder 10.
>
>However i do appreciate the fact that one author is saying the truth about the
>other author.
>
>Any claim of having a lot of knowledge in rybka, which for the average person
>suggests it has more than others, must be proven of course and the obvious proof
>which everyone can see in your assembly code is that it has real little. So it
>was a clear lie. The "advocate of the devil" type of explaining things is not
>relevant.
>
>A claim can be very valid from juridical viewpoint but very wrong from social
>viewpoint.
>
>Chrilly clearly pointed that out for which we have to thank him.
>
>Please in future remember the big difference between lying, juridical non-lying
>and marketing claims. There is a big difference between the two. A good
>dissassembly specialist like Chrilly, his findings there are very valid.
>
>Perhaps take over some marketing slogan from Chessbase a few years ago:
>  "Rybka learns through search".
>
>Nothing wrong with that.
>
>By the way, it wouldn't hurt to show what you do in search, all the top 20
>programmers gonna debug it anyway in your program, the top 2000 however which is
>in this forum who can't read assembly as well as Chrilly nor Frans, they like to
>know :)
>
>Thanks,
>Vincent

It's interesting to see what the non-manipulated KN/s of Rybka was 1.75 years
ago.  Perhaps that project was scrapped shortly thereafter...

http://chessprogramming.org/cccsearch/ccc.php?art_id=356880

http://chessprogramming.org/cccsearch/ccc.php?art_id=356884

-elc.



This page took 0.01 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.