Readers' feedback on the ICGA/Rybka scandal has been published at Chessbase, including Dr. David Levy's response, which includes a link to Dr. Mark Watkins' rebuttal. Kodos to Chessbase for at least doing this.

@sockmonkey Good point - funny how people sign anything to get their prize and publicity but fail to comply with what they signed up for!

Saga continues...

Response to the articles published earlier on Chessbase by ICGA president Dr. David Levy and technical rebuttal by Mark Watkins.

http://www.chessvibes.com/reports/controversy-over-rybkas-disqualification-and-ban-update

Seems also that Chessbase got a copy. I wonder if they'll publish it.

Ultimately, Riis hasn't done anything in this series but rehash a bunch of mostly tangential arguments already found on the relevant computer chess fora:

- Rybka is original because Vasik Rajlich says so
- Any evidence of code copying is irrelevant because code copying is common practice since the availability of strong open-source engines (regardless of licensing concerns)
- The originality clause of the ICGA is hopelessly outmoded and shouldn't be enforced, despite Vas' agreement to it at the time of his entries
- There is no difference between copying code and copying ideas, and copying ideas is fine, or maybe...
- Both copying code and copying ideas are bad, but everyone does it, so what's the problem.
- Rybka was singled out
- The ICGA members (particularly Hyatt) who attempted to explain and/or defend the ICGA ruling are senile freaks with an axe to grind
- Rajlich's genius and ability to improve the code he began with excuse any wrongdoing, because genius should be rewarded, not hampered by inconvenient rules.

All of this stuff is fairly irrelevant, though. The technical evidence makes abundantly clear that the initial development of Rybka was based first on Hyatt's Crafty (open-source, licensed), and then on Letouzey's Fruit (open-source, licensed) in ways a) incompatible with the software licenses and b) incompatible with the originality rule agreed to by the Rybka team when they entered the ICGA competitions.

No one disputes that Vasik Rajlich added a great deal of his own IP and creativity to the software as it developed. Nevertheless, rules are rules, and Rajlich's competitors (both commercially and, more specifically, in the ICGA competitions) were at an unfair disadvantage, playing against an engine whose whose author took shortcuts and whose first 2800 Elo came essentially for free.

This may be acceptable in the business world (although there are still licensing concerns which haven't been resolved), but it's clearly not OK in the context of ICGA tournaments, and they were well within their rights to put a retroactive stop to it. Had Rajlich cooperated with the investigation, provided sample source code to the investigators or other documentation to explain the high number of "coincidental" similarities between Fruit and Rybka (or Crafty and Rybka), the punishment may have been dealt out a bit lighter, but who knows.

Instead, Rajlich's contempt for the proceedings, and contempt for the organization which awarded him the "Computer Chess World Champion" title and prize money, and whose rules he agreed to, weighed heavily against him.

EDIT: The rules that the Rybka team agreed to include a requirement that, in the event of an originality investigation, they provide full source code to assist in the investigation. The very fact that the team refused to cooperate was enough to disqualify the entries, and the fact that the ICGA invested countless man-hours in decompilation, analysis, reconstruction, etc. demonstrates a fair amount of goodwill on the part of the ICGA board.

The results of this work are then dismissed as "fantasy code" by the critics, who conveniently ignore the fact that, had the ICGA board been provided with source code, as stipulated in the competition rules, the entire de- and reconstruction efforts would have been unnecessary. Or do the critics believe that Rajlich gets a free pass if he refuses to cooperate? Unsurprisingly, this point is simply glossed over in Riis' assessment.

Part 4 now published at Chessbase.

I think this is an arguement about Intellectual Property, obviously Rybka has unique IP and would not have been the top engine otherwise, however the GNU liscences also have IP and so perhaps some of the rating belongs to the liscence holders. Basically how many of the other engines use fruit-like search ... in fact, what is fruit-like search anyway? Its very interesting.

The simplest solution for Rybka to prove its innocence would be to make public the source code of any old version of Rybka. If it is choosing to hire a mathematician to argue its case instead of this..... :-)

In part 2 Dr. Riis, a Rybka forum moderator, talks about engine similarity testing. There is another dendrogram by Kai Laskos at TalkChess:
http://talkchess.com/forum/viewtopic.php?topic_view=threads&p=430802&t=40795

Note how similar Fritz is to Rybka and Strelka. Yet, a quick code analysis showed that there is nothing wrong with Fritz. This clearly proved that dendrograms are completely unreliable and could not in no way be served as any kind of proofs.

The 0. explanation in part 3 is just hilarious.

I guess the monkey Mr. Rajlich trained to type his code got drunk and started to press random keys.

When I interned at Microsoft over the summer, I wasn't even allowed to look at another competitor's code, especially open source, for fear of such a controversy.

Part 3 of the chessbase report has now been published.

@browni3141 - Stealing ideas is not illegal, and "stealing" is a loose term. Copyright does not protect ideas, only the tangible expression of those ideas. Rybka may or may not have violated the copyrights of Fruit or Crafty by copying the expression of those ideas by lifting code verbatim. From what I can tell there is a serious possibility it did happen here, and there is a cloud of suspicion, but only a court is equipped to make such a finding conclusively.

We are talking about a research community here. People share ideas with each other and need to trust that they will not see this sharing abused by having others take credit for their contributions as if they are there own. The chess community benefits from better chess engines when this community can reward its participants based on technical merit, and that requires attribution.  Nothing obligated Vasik Rajlich to respect the attribution requirements other than his own entry in a contest whose rules require diligent attribution. By entering, he voluntarily agreed to be held to this standard. He profited from the credibility that winning provides, but he cheated by trying to have it both ways: he wanted all the credit for developing all the ideas within his engine himself, but didn't actually do it. The punishment for winning a contest by breaking the rules is not a legal one, you just have your title revoked and the shame that comes with it, as proscribed by the rules.

Interesting article, finally the other side of the arguement. 

@browni3141

According to ICGA investigation, they did exactly that.  They took the code and used it in their own program. And as I've said ICGA is the only one who can judge on what is defined by using someone elses code. If this was copyright claim on court, it would be extremely difficult for developers of Fruit or Crafty to prove they stole their code, but this is sporting event sanctioned and supervised under ICGA rules, so they alone are the judge.

Again, I think life ban is extremely harsh. If they can make a new engine from ground up, they should be allowed to enter competitions. But there is enough evidence that they used actuall code from Crafty and Fruit.

Whole story is quite ironic, really. Rajlich himself lead quite an crusade against another russian program called Strelka, claiming that it was a copy of Rybka, where in fact that program was also strongly based on Fruit. Naturally they both produced same results. Another proof that Rybka is a clone.

I agree with Dr. Søren Riis 

It think that the stronger argument in his paper is that WCCC rule 2 is not precise. If you are going to be lifetime banned from competition, then I think that it is fair to ask the rules to be clear to a fair degree. And if WCCC rule 2 was precise, then all the questions that are posed in the first part of the paper (under the heading "WCCC rule 2") would have a straightforward answer; ant it is plain clear that they do not have one.

The WCCC should be clear about how to understand rule 2 before applying the rule.

@simonus: From what I understand taking other's ideas is not against the rules, but taking exact code is. There is a big difference between original code and an original idea.

@bwtaylor: You said that Rybka stole ideas, but I was under the impression that it stole exact code, and you can't really 'steal' scientific/mathematical ideas anyways, because no one can truly own them in the first place. The very idea of owning ideas is incredibly discusting to me.

@browni3141

 

There is nothing wrong with modifying and idea. Unless it is forbidden. And in this case it is. If you want to compete in WCCC, you must write an original code. Its as simple as that, and when it was found out that they didn't do that, they were banned.

The engine competition rules require all competitors to give a full disclosure of all researchers whose work was used. Rybka was disqualified because it  implements algorithms taken from both fruit and crafty. Whether or not this rises to the level of copyright infringement is not important to the disqualification, though it is pretty important to Rybka's business model. The fact that the Fruit and Crafty creaters were not properly creditied for their contributions to Rybka deprived them of the recognition they deserved and it's this recognition that the competition's contract was designed to protect. The lack of attribution might be a smoke screen to hide copyright violations as well.

Were Rybka found to have lifted code from Fruit, the entire business model of the Rybka product would be threatened, because Fruit is licensed under the terms of the GPL, which require that all derivitives also be licensed under the GPL, which explicitly allows 3rd party redistribution. In other words, if Rybka is a derivitive of Fruit, you should be able to offer Rybka for open download for free, which obvioulsly impacts their ability to charge for it. Additionally, Rybka clones like Robbolito would go from being unlicensed and infringing (per the Rybka EULA) to licensed (per the Fruit GPL), and this would have a dramatic affect on future iterations of these competitions.

Many people have raise the point that you can't make a stronger engine by copying a weaker one. This is completely naive and doesn't matter at all to either copyright or to the engine competition rules. The presumption is that the free ride taken by lifting code displaced hard work that would have slowed the overall progress. It is doubtful that Rybka would have progressed to the strong ELO range as quickly had Rajlich had to divert his programming attention to figuring out these details himself. Rybka won the competition because of an unfair time to market advantage obtained by stealing ideas that the rules required to be attributed. Attribution is time sensitive. Who knows if the author of Fruit would have asserted his rights more aggressively if he was aware of the use of his code in a more timely fassion. Perhaps he could have obtained more support to continue Fruit from either the recognition or his proper share of the Rybka licensing revenue.

It is not a defense to copyright infringement to show that some parts of the work are original. Rather you have to show that no parts were illicitly copied, and that's what seems to be at issue here.

@etourneau: You said:

"From a moral standpoint, it is wrong :

-to take credit for something you didn't do"

I agree with you there, but what I said is different, and I stand by it:

"How do we know that Rajlich didn't just study the code of Fruit to understand the algorithms it was using, and then improve them? Then obviously the behaviour would be very similar, and credit shouldn't be needed."

You said:

"From a moral standpoint, it is wrong :

...

-to sell something that someone else gave you for free under the explicit condition that you yourself give it for free to others"

I agree with this in the case of the original work, but not in the case of a derivative work. A derivative work is in my eyes something completely new, and this statement should not apply to such a work. Of course it is hard to know where to draw the line. If you take a program and change a single bit in the machine code, that hardly makes it something new, but certainly taking someone elses code and modifying 50% of it does make it something new.

If you still think that copying exact code is unethical then what about taking someone elses idea, implementing and modifying it. Surely that is enough to make it your own, and credit wouldn't need to be given.

Hilariously, Dr. Søren Riis seems to be a moderator at the Rybka Forum! Certainly not remotely independent at all!

In my mind there is nothing that Chessbase can do about the ban. ICGA has same legal status as FIFA, UEFA, IOC... it has its own rules and is not governed by calssical legal system. Long letters from doctor of science (notice that he isn't a lawyer) have no meaning here.

As for the ban itself, I think it was justified, but also too harsh.