Rybka disqualified and banned from World Computer Chess Championships | Hacker News Reader

Rybka disqualified and banned from World Computer Chess Championships

chessvibes.com

57 pointsfogus15 comments

I haven't followed computer chess for a few years, so wasn't aware there was such a growing controversy, especially considering how dominant Rybka has been(or was).

Maybe someone who is more familiar with the case can explain what happened to warrant such an action?

Crafy(free/open source though not GPL) by http://en.wikipedia.org/wiki/Robert_Hyatt has been one of the cornerstones in computer chess.

I am sure most aspiring computer chess programmers have looked at its source and techniques.

Fruit I also believe was open source(not sure what licence).

Now, both Crafty and Fruit are very strong GM level programs on a modern computer but are far from being the best. In fact, I(a lowly FM) routinely was able to find flaws in their algorithms (especially Crafty) and draw or win games using standard tricks(closed openings, passive no obvious weakness openings, etc.).

Rybka is pretty much immune to such anti computer chess shenigans. It has a much better "knowledge" of human chess somewhat explained by fact that Rajlich is an IM.

In fact, Rybka stood heads and shoulders above all the other contenders for a while.

There have been rumblings for years, at least since Strelka came out. I happen to think the GPLed code of Fruit was the best thing to happen to computer chess for a long while (thank you to Fabian). IMHO, cloning is not a bad thing since allows incremental improvement by multiple authors. Cloning without respecting the license of the source, however, it not okay and it seems clear now that's what Rajlich did.

Thank you for the explanation!

I took the time to read the official report on Rybka Investigation and Summary of Findings for the ICGA http://www.chessvibes.com/plaatjes/rybkaevidence/RybkaInvest...

For those interested in the issue, I recommend you read the report, as well.

Some real heavies weight in on the issue including Ken Thompson. All are in agreement that early versions of Rybka show disturbing similarities to Crafty and then Fruit(this is when the big jump in rating performance was made).

Rajlich has continued to claim that his code is all original which is highly doubtful.

To me the real damning evidence was by Hyatt, which showed that the early version of Rybka has whole bunch of useless code from Crafty, which Hyatt had used:

"Summary: Pre-Rybka 1.6.1 contains much identical code to Crafty, even including large blocks of code with obsolete code inside them, and code that performs tests that make no sense today (code that was left in Crafty by accident, by Robert Hyatt, also shows up in Rybka 1.6.1)."

Off course, most of the code must be gone by the later versions, but it is seriously in such a bad taste to be so disrespectful to other programmers.

I haven't seen the evidence, but it seems to me to accuse him of code theft because the moves end up being the same seems a bit much.

Did he refuse to show his source code?

Also, claiming that he knew he was in the wrong because he denies being in the wrong is too much, too. He's more guilty because he claims to be innocent? Jeez.

Also, demanding the trophies and prize money back is fine, but if they take him to court over them, that's going too far. He might very well not have the money any more. That it took them 6 years to figure out he was cheating says it wasn't very obvious, to me. And now they claim that there is overwhelming evidence.

I dunno, something smells fishy to me. I'd certainly want more proof than 'it does the same moves'.

"demands that he returns the trophies and prize money of the World Computer Chess Championships in 2007, 2008, 2009 and 2010."

What if he just says "no". Is there anything legally compelling him to comply?

If they have a good case for plaigarism, I'm sure they could successfully sue him for that money, and then some.

But will they? It looks very bad and lawsuits cost a lot of money.

I don't think it looks bad; it helps to maintain the integrity of the tournament, encouraging fair contestants.

I'm a little confused about what he's been accused of here. Are they saying he cut and pasted code from other chess bots? Or that he examined other chess bots and used the ideas they implemented? Or...? What exactly do they mean by plagiarism?

Essentially direct source code copying.

The attached document indicates that Rybka contains the exact logic that is supposedly uniquely implemented in either Crafty or Fruit. These are not standard algorithms, and the likelihood of a closed implementation of a chess engine exactly matching the logic of an open sourced version on a number of unusual algorithms cannot be down to pure coincidence.

I guess the developer behind Rybka studied the code behind both Crafty and Fruit, and elected to use pieces of code from them, but didn't acknowledge/give credit for those pieces of code.

One of the example pieces of code is a hack to avoid enpassant specifically in a King and pawn versus a King and pawn endgame - something rather unusual considering the same issue occurs if either/both sides have more than 1 pawn. Fruit's approach was just to treat the KPvKP, and so to did Rybkas.

Again, the algorithms documeted were based around unique approaches to evaluating various facets of evaluating the score of a position, not something as simple as "can the White king castle in this position". So they don't make a massive difference in playing strength, which suggests more strongly that there's been some uncredited copying going on.

Rybka has been the strongest rated chess engine for a few years now. Granted that the developer has found definite improvements over the other chess engines out there. But, for the purposes of World Computer Chess Champion, the developer must own the program he runs (or document/credit the sections of the code he doesn't own)

Definitely have a look through the evidence. It's quite surprising how detailed it is. Comparing Crafty/Fruit source code to a decompiled Rybka codebase and looking at specific unusual engine-specific edge-cases and exceptions.

Let this be a lesson, kids: cheaters never win.

While I'd love it if it were true, that's not a valid conclusion. How many cases of winning cheaters are out there that never got caught?

Only "cheaters that get caught never win" would be true, but that's kinda obvious.

I know. I meant it somewhat ironically. ;P

I wish this was the case here, but Rybka has clearly benefited financially its "creator"(quotes because of this uncredited copying).

As the strongest chess engine in the world for 3-4 years, it must have sold quite a few copies.

By comparison, Crafty and Fruit are free and thus unlikely to bring much in a way of a financial windfall.

The big difference between this case and countless others where cheaters benefit financially (think about thinly disguised GUI front-ends for FFmpeg, which do not mention GPL anywhere), is that Rybka truly was a superior product from those it copied.

Hard to speculate on the motives here but Rajlich must not have wished to share his laurels with anyone.

Reply on news.ycombinator.com