Rybka disqualified and banned from World Computer Chess Cham

[Albert Silver]

For now ChessBase is still selling Rybka.

clearly, he can keep selling it.

wouldn't further action would be needed, i.e. an expensive and victorious law suit,
brought by...FSF or Ryan Benitez, or Bob Hyatt for damages...hmmm
loss of income? no i guess not

well who knows?

this however is simply a disqualification from a gaming association, a rather harsh 'slap' on the fanny, and go stand in the corner,
(thanks John Conway for the nice image and apt analogy)

sorry, but it probably won't faze them one bit IMO, and it's clearly a product for which there's a demand.

I noticed in the investigation report that two people from ChessBase were also in the panel:

Albert Silver (software designer for Chess Assistant (1999-2002); currently editor of Chessbase News (2010-present))

Frederic Friedel (Chessbase.com)

It seems they were at least interested in the outcome, what they'll do I don't know.

Neither of us were on the panel. We were merely given access to the forums where the debate and analysis took place.

[sje]

Cloning that code, or any other chess-specific code, is nothing like copying a call to a general system library routine.

I don't see it as very different

Until the day comes when there's a Unix system call to probe a TB file, then it is very, very different.

[Shaun]

A sad day for computerchess?

Perhaps it's not the best day for computer chess, but it can be a good day if the actions taken serve as a positive influence on current and future authors who might otherwise might think that they can easily get away with unauthorized cloning.

The treatment of Vas is not consistent - lets now examine a few more top engines. I wonder how many would pass, who knows a few unknown authors might end up with a world title

If we can come up with a view on what can be copied / used and apply that consistently then perhaps something good will come of this mess.

Even if we assume Vas started from Fruit 2.1, something that I am not convinced of, he added over 450 elo (CCRL 40/40) over 6 years or so. This is not the same as taking an open source engine or decompiled source and making some tweaks...

Shaun

[michiguel]

For now ChessBase is still selling Rybka.

clearly, he can keep selling it.

wouldn't further action would be needed, i.e. an expensive and victorious law suit,
brought by...FSF or Ryan Benitez, or Bob Hyatt for damages...hmmm
loss of income? no i guess not

well who knows?

this however is simply a disqualification from a gaming association, a rather harsh 'slap' on the fanny, and go stand in the corner,
(thanks John Conway for the nice image and apt analogy)

sorry, but it probably won't faze them one bit IMO, and it's clearly a product for which there's a demand.

I noticed in the investigation report that two people from ChessBase were also in the panel:

Albert Silver (software designer for Chess Assistant (1999-2002); currently editor of Chessbase News (2010-present))

Frederic Friedel (Chessbase.com)

It seems they were at least interested in the outcome, what they'll do I don't know.

Neither of us were on the panel. We were merely given access to the forums where the debate and analysis took place.

Then why are they using your name? did you sign or agree on the final report?

Miguel

[michiguel]

Cloning that code, or any other chess-specific code, is nothing like copying a call to a general system library routine.

I don't see it as very different

Until the day comes when there's a Unix system call to probe a TB file, then it is very, very different.

Just for the record, the day in which TBs could be probed using a call to linux/unix library has come.

Miguel

[menniepals]

The last Rybka engine I bought was Rybka 3. But to support the team, I will buy Rybka engines whenever a new one is release from now on. The events that transpired lately should not discourage me on anything about Rybka. It still remains to be the strongest and the best engine in the planet. Fruit and Crafty should not have released their sources in the first place if they don't want people learning from them. I stand firmly convinced, this is unjust for the Rybka team.

[Shaun]

Cloning that code, or any other chess-specific code, is nothing like copying a call to a general system library routine.

I don't see it as very different

Until the day comes when there's a Unix system call to probe a TB file, then it is very, very different.

Just for the record, the day in which TBs could be probed using a call to linux/unix library has come.

Miguel

[sje]

Cloning that code, or any other chess-specific code, is nothing like copying a call to a general system library routine.

I don't see it as very different

Until the day comes when there's a Unix system call to probe a TB file, then it is very, very different.

Just for the record, the day in which TBs could be probed using a call to linux/unix library has come.

Gosh, I must have missed the documentation update for I can't find the man page for

Code: Select all

score = tblookup&#40;<FEN-string>);

Maybe it's in the same section as

Code: Select all

move = ab_pvs_search&#40;<FEN-string>, <depth>);

[wgarvin]

It is not so easy to make something that is more than 100 elo better than other engines even without moral values and I think that it is easier to get a 2800 engine without copy/paste from other sources.

I agree at least with the first part of this - I think the Rybka team made significant independent contributions to their engine and for this reason I believe they deserve some credit.

There is no doubt that the Rybka team wrote a lot of original code and made Rybka a very strong engine.

However, there is also evidence of some plagarism from Crafty and Fruit in some versions (almost the entire eval from Fruit 2.1 was copied and ported to bitboards around the time of Rybka 1.0 beta, and there is evidence of Crafty code in pre-Beta versions of Rybka). The ICGA investigation needed to decide if Rybka had violated the tournament rules of any ICGA events it participated in. One of the rules states that all of the game-playing code in the program must be the "original work" of the authors who are entering it in the tournament, and not "derived from or including game-playing code written by others". The names of all authors have to be listed on the entry form. For years, Rajlich maintained that Rybka was original work, and evaded questions about its relationship to Fruit. He entered it in several ICGA tournaments without acknowledging that Rybka contained an evaluation which was "derived from or including game-playing code written by others".

The WCCC and other competitions are a chance for the world's best chess programmers to compete against each other with their creations. But everyone must write their own program -- if they were allowed to copy significant parts of other people's programs into their own, without attribution, that would give them an unfair advantage. I believe that by copying Fruit 2.1's eval (which was a well-structured eval from one of the strongest open-source programs available at that time), Rajlich saved himself months of hard work. He then entered his program in the WCCC for several years, and other tournaments, without acknowledging that his eval was derived from other people's code.

[bob]

Thumb up 2 Thumb down 1

If the Rybka=Fruit/Crafty connection is so strong and the evidence so damning, then why did this take over 5 years to unearth? Five and a half years since December 2005! It should be noted , for whatever it’s worth , that a huge majority of the panel members who investigated and condemned Rajlich are his direct competitors, ie, authors of rival programs. These same competitors wonder why Rajlich did not give up the Rybka source code for their perusal.

As a layperson in computer matters (I am an attorney), I don’t get this at all. There is no Rybka source code to compare, so apparently various circumstantial evidence is cited. Fruit was the strongest engine. Then Rybka was the strongest engine. And yet their evaluation functions were found to overlap by 64%. Call the police! The panel points out that weaker engines had less overlap as compared to the 2 strongest engines (Fruit and Rybka). Naturally, engines that are closer to the “chess truth” will overlap more. I don’t see how an evaluation overlap of 64% is “nearly identical”. Geneticists claim that human beings share 96% of the same DNA (aka “source code”) with chimpanzees. Perhaps the ICGA will ban God from claiming that Human Beings are uniquely in His image since the ALmighty seems to have plagiarized the source code for chimpanzees when coding humans..

[Reply]

The similarity in search and eval are not very condemnatory in my view.

The duplication of errors and dead code is.

I do not think that it has been demonstrated that Rybka has broken the law. But it seems to me that their finding is not unreasonable as far as breaking the conditions of the contest. While he does cite the named programs as sources of information, I find that things like initializations that only have purpose for Steven J. Edwards tablebase files being duplicated should make everyone uneasy about it.

IMO-YMMV

I downloaded all the documents and printed them. I am reading them now.

The code in question is _not_ about "edwards table initialization". It is about code in iterate.c that works around a known bug in the edwards files that Crafty had to deal with, but Rybka did not. It has to do with code in evaluate.c (crafty) that was a clear example of code that was tested, then removed, but not quite "all" of it was removed, leaving some code inline that could never be executed, and code which no normal person would write (by itself) since it was added because of other test code that was later removed from Crafty.

And, of course, there is the issue of other pieces of Crafty, all should now be public in the ICGA Wiki so that everything we considered can be examined publicly.