Computer chess: Does it matter?

According to an article at ChessVibes, the International Computer Games Association has disqualified Rybka from the present and all future World Computer Chess Championships, and has also voided its world titles from 2007-2010. The reason is that Rybka was deemed, by a five-man panel of judges, to have plagiarized from two open-source programs, Fruit and Crafty.

The legal and computer science issues are head-spinningly complex, and I can’t really presume to judge who is right or wrong in this matter. To be honest, I had never even heard of Fruit until this morning. For readers that want to find out more about those issues, please go to the ChessVibes article and read the long and informative comments thread. Let’s just cut to the chase: Does it matter for chess players?

I’d like to hear what readers of this blog think, but my feeling is that it doesn’t matter very much. So Rybka can no longer call itself a world champion chess program — but everybody still knows that it is incredibly strong, and people will continue to buy it and use it. Some people might say that it’s moot because Rybka is no longer the strongest in the world, as of 2011. A free program called Houdini beat Rybka 4 earlier this year in what might be called a non-official world computer championship. Also, some of the ChessVibes commentators say that Stockfish is now better than Rybka.

But something funny is happening. I’m starting to wonder whether a proprietary chess program can ever be world champion again. To get to the top of world computer chess, you need to borrow so many ideas from the programs and programmers who have gone before you. The authors of these open-source programs allow you to borrow your code, as long as you also share yours. You can’t jump on the open-source bandwagon halfway and then jump off once your program gets to be better than the previous ones. That’s apparently what the creator of Rybka, Vasik Rajlich, tried to do.

That’s why you’ll never see a commercial version of Houdini (at least if I understand the matter correctly). Houdini borrows all over the place from other programs, such as Ippolit and Rybka itself.

For advocates of the open-source movement I think that this news will be very good. We will have two kinds of chess software in the world, the inferior commercial programs and the superior free programs. As someone who is not a techie, though, I’m not sure whether this is a positive development. I tend to distrust free software because it has a clunkier user interface and I worry that it may have bugs. (Not that commercial software is bug-free, but at least they have huge teams of engineers looking for and fixing the bugs.) Like many people, I don’t want to commute to work in some mad professor’s concept car. One reason I am willing to pay for software is that the brand name is a seal of quality.

So there is some sense in which it’s demoralizing to realize that the concept cars will always win, and that I really can’t have the world’s best chess software on my desktop. But how realistic was that desire, anyway? I’ve long ago accepted that I can’t have the world’s fastest car in my driveway. (Or the world’s most fuel-efficient car, which would be more interesting.)

And again, I come back to the question: Does it matter? Rybka is good enough for any human use — preparing your openings, going over your tournament games and looking for the outright blunders. In fact, even Fritz is good enough for human use. It’s fallen out of fashion because it’s no longer the world champion, but it’s still like having a grandmaster at your fingertips.

Instead of running out to buy ever-stronger chess programs, consumers like me might actually benefit more from a weaker chess program: one that has been programmed to downgrade variations that are not human-understandable. There are so many times that I get sucked into computer analysis — sometimes even beautiful computer analysis — but at the end of it I have to shake my head and say, “But could any of that ever happen in a human game?”

I would like to offer this as a challenge to future chess programmers: Can you write a program that will offer not only an “Analysis mode,” but a “Human analysis mode” that would evaluate positions and variations using more human-like criteria? If a particular variation requires you to find an incredibly precise series of oddball moves in order to get an advantage, the “Human analysis mode” would flag this as an unreliable variation. Or if I play a sacrifice that can only be refuted by some extraordinary trick that no human would see, I would like to have the sacrifice flagged as “Sound enough for all practical purposes.” Can you do that? Pretty please? I think that would be a bigger service to human chess players than coming up with the next Houdini or Stockfish.

P.S. There’s one other interesting aspect of this story that some of the commentors on ChessVibes point out: the ICGA sanctions against Rybka have so far gone completely unreported by Chessbase. One has to wonder why. Is it because Chessbase still wants to be able to sell Rybka software? If so, it suggests that Chessbase will put business interests ahead of journalistic interests, when push comes to shove. That’s kind of disturbing if, like me, you have come to think of Chessbase as your go-to site for international chess news.