The rumors about Strelka beeing a Rybka clone is still alive

[Dariusz Orzechowski]

It would be interesting if Juri Osipov released the source code of Strelka. I wonder how many of morality pseudo-defendants would refrain from taking a look.

[Sylwy]

The source code of Strelka are already in possession of Sergei Markoff,Dann Corbit & Bryan Hoffmann .
Sylwy

[Uri Blass]

I am reading this discussion for a few days now and I am quite surprised that it doesn't end.

If I remember well, there were rumours, a few years ago, that Rybka too was a Fruit clone, and Vaclav Rajlich, in an interview found on the Alex Schmidt site, admitted he had used some ideas from the Fruit code. (There is something interesting in the fact that Rybka, although tested previously, has made his way only a few months after Letouzey's work.) But he was obviously not the only one. A lot of engines made a tremendous jump in strength after Fruit (an average of more than 60 elo): HIARCS, Fritz, Loop, among others. Just look at the CSS ratings. The only one not to improve dramatically was SMK's Shredder. Did anybody accuse their authors? No. They found new ideas (in the code of Fruit, perhaps) and they implemented them.

Suddenly there is a young Russian with a new engine: the story is starting all over again, and this time, it is an incredible one. Of course, Osipov has probably read all the Ed Shroeder's articles about programming, and, of course, he reviewed Crafty's, Toga's, Fruit's and Glaurung's code. It is natural that he also tried to discover Rybka's secrets, to have a similar evaluation function, for it is an excellent evaluation function and everybody is doing it too (Christophe Théron has admitted it in this same forum). Does it make him an immoral person, an intellectual thief? One of us has pointed out that no one is starting from the beginning, and that there isn't any engine that is absolutely original. It is so comprehensible.

Then, how could he have stolen Rybka's code? We can see that Vasik Rajlich is doing all he can to protect his code: faking the node count, the size of the -exe, encrypting the exe... If Osipov was German or American, the allegations of some of us could have lead to court: to accuse somebody that he is a (the) hacker that stole code from Vasik Rajlich's computer without proof is a felony. And the premise of the law system in the majority of countries of the world is that someone is innocent until the contrary is proven.

Finally, there are the affirmations of Dann Corbit, Bryan Hofmann and Sergei Markoff – the only one who have indeed seen Strelka’s code, that Strelka is not a clone. Of course, they didn’t see Rybka’s code, but nobody claiming Strelka is a clone has seen Rybka’s code neither. The solution for me is clear: Vasik Rajlich could send his code to Brian Hofmann and Dann Corbit to compare. Why not. (By the way, they could see if Rajlich has changed Fruit search function or has integrated it in his code without changes. ) Osipov, to clear his name, has done it. For me, Vaclav Rajlich's silence about this subject could indicate something else: Osipov has found something Vas wouldn't like to be discovered by others, an idea or the idea that makes Rybka's so strong.

Dann Corbit has made a judicious effort to clear the notions of clone, reverse engineering, what is legal and what is not, and to put an end to this discussion. He is perfectly right: we should be ashamed. Let us hope there will be further versions of Strelka.

This is simply not correct that everybody is doing it too.
The problem is that osipov copied the also the clearly wrong part of rybka's evaluation so it does not seem that he tried to learn from rybka's evaluation to improve his evaluation function.

king vs king should be evaluated as 0.00
I do not know of a single program that does not evaluate king against king as 0.00 except rybka beta and strelka.

2 programmers may have the same bug or lack of knowledge in one case but the fact that they have exactly the same evaluations in king vs king and not only not having 0.00 is telling us that it is probably a clone.

one same bug may be an accident but I could find more bugs that both strelka and rybka have and in KQ vs KPPP the score for the stronger side with the queen goes down significantly relative to the initial score only for 2 programs(rybka and strelka)

Uri

[Michael Sherwin]

Hi Uri,

Could both Rybka and Strelka be based on the same open source program from Russia or another eastern European country that the world does not know about? Maybe if such a program exist, it may not be winboard or uci compatible. Other than open source programs in English and some in Spanish, I do not know of any that are in a different language. Surely there must be doesens of sources available in eastern European countries that are not in english!

Mike

[Guetti]

Hi Uri,

Could both Rybka and Strelka be based on the same open source program from Russia ....
Mike

Kaissa!!!

[Alexander Schmidt]

Strelka is not a clone.

Having the same evaluation as Rybka for some positions is not a crime as far as I know.

Hello Dan,

http://www.talkchess.com/forum/viewtopi ... 91&t=14925

We have exactly the same evaluation in all depth in all KvsK positions, out of that we have nearly exactly the same UCI output which is usually different in different engines.

So far I defended Strelka, but this is an evidence that there is something in common between Strelka and Rybka.

That the Strelka Code looks unique is fine, so Rybka seems to be a unique engine.

Best,
Alex

[K I Hyams]

one of his testers got access to his computer.

I had already decided that this was the most likely possibility. One of the testers of the early versions is known to have a mischievous streak............

[Norm Pollock]

one of his testers got access to his computer.

I had already decided that this was the most likely possibility. One of the testers of the early versions is known to have a mischievous streak............

The butler did it!

[mjlef]

LOL

Vasik just posted something on the Rybka forum you should read:

http://rybkaforum.net/cgi-bin/rybkaforu ... id=1655#fp

he says "1) There is no doubt in my mind that the similarities (in executables and in output) between Rybka 1.0 and Strelka 1.8 are not a coincidence. This is in contrast to Strelka 1.0, where the similarities didn't seem to me to be excessive."

He also casts doubt of anyone being able to steal RYbka, stating it has only been on one machine and never connected to the internet (he needs to watch Mission Impossible to see how those clever guys can get anything!)

Interesting. Has anyone reviewed the source code of Strelka 1.8, or just 1.0?

Mark

[Tony Thomas]

Hi Uri,

Could both Rybka and Strelka be based on the same open source program from Russia ....
Mike

Kaissa!!!

kaissa, now which one are you talking about? The famous one or the infamous one?