Rybka disqualified and banned from WCCC

[Jeremy Bernstein]

I have to admit: I am surprised by the severity of the punishment. I haven't read all of the evidence yet, but even given the old documents, which I have read, the ruling seems, if not just (which is probably is), at least justified. Whichever way this issue cuts for you, it's not happy news. Nevertheless, it's a triumph for ethics in the narrow confines of the CC programming field, and the ICGA (and its investigative committee) should be applauded for their willingness to take the high road.

http://www.chessvibes.com/reports/rybka ... pionships/

[ThinkingALot]

Looks like ICGA just banned itself from using Rybka:).

I wonder how they are going to prove that Vasik used 1.0 beta (the only Rybka version accused of being a clone) in their tourney... Otherwise the claim that he played with a Fruit/Crafty clone has zero evidence behind it.

[mjlef]

I suggest everyone read the evidence presented. Analysis was made of pre-Rybak 1.0 beta, Rybka 1.0 beta and versions up to Rybka 2.3.2a, to "straddle" any versions of Rybka used in the early tournaments. The analysis is quite comprehensive and should be reviewed before people reach conclusions.

[ThinkingALot]

Looks like ICGA just banned itself from using Rybka:).

I wonder how they are going to prove that Vasik used 1.0 beta (the only Rybka version accused of being a clone) in their tourney... Otherwise the claim that he played with a Fruit/Crafty clone has zero evidence behind it.

EDIT: I looked through the "evidence documents". It appears that a thorough analysis of 2.3.2a was also conducted. Perhaps I missed smth but I still can't see any connection between 1.x-2.x.x Rybka versions and those engines which played as 'Rybka" in ICGA's toruney...

[Jeremy Bernstein]

Looks like ICGA just banned itself from using Rybka:).

I wonder how they are going to prove that Vasik used 1.0 beta (the only Rybka version accused of being a clone) in their tourney... Otherwise the claim that he played with a Fruit/Crafty clone has zero evidence behind it.

EDIT: I looked through the "evidence documents". It appears that a thorough analysis of 2.3.2a was also conducted. Perhaps I missed smth but I still can't see any connection between 1.x-2.x.x Rybka versions and those engines which played as 'Rybka" in ICGA's toruney...

Is this a kind of "you can never step in the same river twice" argument? Which engine do you suppose was playing? Vas' secret private engine based on completely different source code than the engine we normalos call Rybka? I think that the ICGA has demonstrated a pattern of behavior as well as concrete examples of code copying, to the extent possible using RE techniques, for versions of Rybka which were extant in the time frame of ICGA tournaments. If the version in the tournament was 2.2.1f_private, an analysis of 2.3.2a which demonstrates continuity in copied code between 1.x and 2.3.2a is clearly adequate.

Jeremy

[ThinkingALot]

Sure, it's quite unlikely that Vas played with Rybka 3 beta in 2005. However, IMHO, this tiny possibility makes ICGA's demands of trophies and prize money back a bit premature, don't you think? Because such demands are based not on evidence but just on a suspicion.

[karakaniec]

More on my page:
http://szachmaty.blogspot.com/2011/06/s ... owana.html

[BB+]

I too agree the verdict was rather harsh (other panel members had suggested lighter penalties, though in any event the Panel would not be involved in such matters). To quote Levy (from the verdict): The ICGA regards Vasik Rajlich’s violation of the abovementioned rule as the most serious offence that a chess programmer and ICGA member can commit with respect to his peers and to the ICGA. During the course of the investigation and upon presentation of the Secretariat’s report Vasik Rajlich did not offer, despite repeated invitations from the ICGA to do so, any kind of defence to the allegations, or to the evidence, or to the Secretariat’s report, other than to claim in an e-mail to myself [...] When a person doesn't (or is unwilling) to respond to quite detailed allegations about "the most serious offence" that can be committed, it is hard to muster up much sympathy in his defence. To wit: why should I care about what the ICGA does to him, when he himself seems rather ambivalent about it all?

I think what swayed the Executive Board was the general lack of communication and/or respect from Rajlich regarding the whole issue [E.g., regarding a previous comment, it shouldn't be the ICGA that has to waste its time to ascertain out what Rybka version (likely) played in what event, when Rajlich could answer the question rather easily -- but even such basic data-points seem not to have been addressed by him]. From what I could tell (and the Panel was mostly out of the loop for the last 2 months, with only a brief about the Levy-Rajlich emails), it seems that Rajlich didn't spend more than maybe 15 minutes of his time to address the questions and evidence --- when contrasted to the hundreds of hours spent by myself, Zach, and others to accumulate the evidence [not to mention even the 5-10 hours per Panel member needed just to read, digest, and comment on it], his one-sentence response [that Rybka in his view was not "derived from game-playing code written by others"] was essentially construed as a non-defence (or alternatively, almost a challenge as to whether the ICGA was competent to interpret its own rules [cf. in this regard my comment below about the ICGA being more vigilant in requiring more complete descriptions, in which case the question of rule-interpretation could become almost moot]).

If Rajlich had been more communicative, my guess is that there would have been more prospect for a "deal" being worked out with which Fruit could be considered a joint winner for 2007, etc. [viewing the 2008-10 entries as only being "minor" violations, akin to filling out the entry form incompletely]. But I surmise that the Board concluded that it really had no other option [a few months back, I had suggested privately to Mark Lefler that if the proceedings were de facto to be in absentia (not to mention ex post facto), one option would be to first make a declaration, with an ultimatum that if the situation was not addressed by Rajlich by the next WCCC, then more substantial action would be taken -- but I can certainly appreciate that Levy and the Board did not want things to drag on, and so opted for finality at this stage].

The most productive thing I can say at this time (I mentioned the same in a recent email to Fabien) is that I would suggest that the ICGA to be much more vigilant in requiring adequate "descriptions" of entries [as was more typical prior to the last decade]. No more one-line "X is a program to play Y that I wrote in my free time over the last Z years, combining well-known techniques with some of my own ideas" [well, actually, that's more complete than some I've seen], but (to use what I could expect for Rybka in 2006): Rybka is a chess engine I have been working on since 2003. It originally started as a bitboard MTD(f) salvaged (and then refitted) from some Crafty parts, but I switched to PVS last year. Fruit 2.1 had a big influence too, and after it appeared I largely rewrote my evaluation function and tuned its "minimalist" feature set. [Fruit also helped me clean up my UCI parsing and time managment code]. Special aspects of Rybka include a "material-imbalance" table that I've been playing around with since the beginning, and some new ideas I've had about search involving pruning and LMR. Rybka went commercial in late 2005 in a single-cpu version, and much of my recent work has been to add parallel SMP capabilities.

I also must say that I don't necessarily agree with such things as "promoting" the 2nd-4th place finishers, but having a "vacated" title is no panacea either.

Well, onto the Loop 2007 case...

[Matthias Gemuh]

Sure, it's quite unlikely that Vas played with Rybka 3 beta in 2005. However, IMHO, this tiny possibility makes ICGA's demands of trophies and prize money back a bit premature, don't you think? Because such demands are based not on evidence but just on a suspicion.

Even Vas himself does not claim to have played with Rybka 3 beta in 2005.

.

[hyatt]

Looks like ICGA just banned itself from using Rybka:).

I wonder how they are going to prove that Vasik used 1.0 beta (the only Rybka version accused of being a clone) in their tourney... Otherwise the claim that he played with a Fruit/Crafty clone has zero evidence behind it.

What are you talking about? The report goes far beyond 1.0 beta in its analysis and conclusions...