Sensational! Y.Ossipov talking about STRELKA

[Andrej Sidorov]

Vasik Rajlich told:

I went through the Fruit 2.1 source code forwards and backwards and took many things.

http://www.superchessengine.com/vasik_rajlich.htm

So Osipov no more guilty than Rajlich.

[GS]

Vasik Rajlich told:

I went through the Fruit 2.1 source code forwards and backwards and took many things.

http://www.superchessengine.com/vasik_rajlich.htm

So Osipov no more guilty than Rajlich.

This is simply bad mouthing or a language problem. Read the _whole_
interview again and may be you can grasp this time that he meant the ideas
of course. He did not 'took' any code but the ideas of Fruits algorithms.

Guenther

[Onno Garms]

But nowaday projects start from Fruit, Toga, Crafty, and after a bit of coding, they are entered into the WCCC (see Gridchess). And nobody objects or cares about GPL.

The GPL does not force you to release your modified copy. So playing an engine that is based on GPL code in a tournarment does not violate the GPL. The GPL only states: If you release, you have to release the source.

If tournament statutes allow engines that are based on others is the tournament's organizers decission, not a matter of the GPL.

[Andrej Sidorov]

He did not 'took' any code but the ideas of Fruits algorithms.

Guenther

So didn't Osipov. Corbit an Hoffman didn't see any resemblance with Fruit in Strelka sources.

[slobo]

I suppose it all depends on how the program was structured. Tord went from mailbox to bitboard yet all the specifics are well placed in functions. So the search part of the code and even the evaluation looks remarkably similiar.

I am still impressed by this reverse engineering job. But I feel bad. SHould I be impressed by a copy of the Mona Lisa?

sorry,
which one is "mona lisa" is this case:
fruit or rybka ?

hf slobo

[Dr.Ex]

"I was only interested in search heuristics and evaluation function. This allowed me to achieve greater similarity (resemblance) with Rybka. In addition, I found that Vasik walked the same way (path) – he has taken Fruit as a basis, rewrote it to bitboard and included tables of material imbalance of Kaufman. Some of the tables I kidnapped (ravished) from the Rybka. "

Wel if that were true, at least Vas had rewritten Fruit.
The same can hardly be said about Ossipov.
I mean, Strelka is so similar to Rybka Beta, even the "Bugs" are the same.

At the time of the Rybka Beta release I discovered what I thought is a Bug in Rybka Beta. If you take away two minor pieces from one side in a material equal position, Rybka showed a very unique behaviour.

For example the starting position without the two white bishops:

rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RN1QK1NR w KQkq - 0 1

Analysis by Rybka 1.0 Beta 32-bit:

1.Sc3
-+ (-3.19) Tiefe: 3 00:00:00
1.Sc3
-+ (-3.34) Tiefe: 4 00:00:00
1.Sc3 Sc6
-+ (-7.36) Tiefe: 5 00:00:00
1.Sc3 Sc6 2.Sf3
-+ (-7.30) Tiefe: 6 00:00:00 194kN
1.Sc3 Sc6 2.Sf3 Sf6
-+ (-7.36) Tiefe: 7 00:00:00 326kN
1.Sc3 Sc6 2.Sf3 Sf6 3.0-0
-+ (-7.30) Tiefe: 8 00:00:03 1753kN
1.Sc3 Sc6 2.Sf3 Sf6 3.0-0 d6
-+ (-7.36) Tiefe: 9 00:00:06 3768kN

And now Strelka on the same position:

Analysis by Strelka 1.8 UCI:

1.Sc3
-+ (-3.06) Tiefe: 1 00:00:00
1.Sc3 Sc6
-+ (-3.43) Tiefe: 2 00:00:00
1.Sc3 Sc6 2.Sf3
-+ (-3.05) Tiefe: 3 00:00:00
1.Sc3 Sc6 2.Sf3 Sf6
-+ (-3.42) Tiefe: 4 00:00:00
1.Sc3 Sc6 2.Sf3 Sf6 3.0-0
-+ (-3.19) Tiefe: 5 00:00:00
1.Sc3 Sc6 2.Sf3 Sf6 3.0-0 d6
-+ (-3.34) Tiefe: 6 00:00:00 2kN
1.Sc3 Sc6 2.Sf3 Sf6 3.0-0 d6 4.Sa4
-+ (-7.36) Tiefe: 7 00:00:00 22kN
1.Sc3 Sc6 2.Sf3 Sf6 3.0-0 d6 4.Sa4 Sb4
-+ (-7.30) Tiefe: 8 00:00:00 724kN
1.Sc3 Sc6 2.Sf3 Sf6 3.0-0 d6 4.Sa4 Sb4 5.g3
-+ (-7.36) Tiefe: 9 00:00:01 1233kN
1.Sc3 Sc6 2.Sf3 Sf6 3.0-0 d6 4.Sa4 Sb4 5.g3 b5
-+ (-7.30) Tiefe: 10 00:00:03 7080kN

So Strelka and Rybka Beta (if Strelka depth - 2 = Rybka depth) discover at ply 7 that they are 2 pieces up and not just one?

[George Tsavdaris]

Vasik Rajlich told:

I went through the Fruit 2.1 source code forwards and backwards and took many things.

This is obvious that means: Ideas and not the code of Fruit.

He(Vasik Rajlich) also said:
"Yes, the publication of Fruit 2.1 was huge. Look at how many engines took a massive jump in its wake: Rybka, Hiarcs, Fritz, Zappa, Spike, List, and so on."

He is honest enough to put Rybka in there. And that as i previously said means he has used ideas there are on Fruit. Not the code..... He said.

He did not 'took' any code but the ideas of Fruits algorithms.

Guenther

So didn't Osipov. Corbit an Hoffman didn't see any resemblance with Fruit in Strelka sources.

Here we are playing with words and actually who knows what he meant in the original Russian post. The translation says:

"1.As the basis of program was taken source code of Fruit (little less than two years ago).
2.The algorithm has been completely rewritten to bitboards. This operation had not changed anything, only improved performance twice (x2). "

From this and from the important "This operation had not changed anything" means he has copied Fruit's source and rewritten it to use bitboards.
Don't know if it's a GPL violation or not, but it's completely different from what Vasik says he has done by seeing the ideas there are on Fruit....

Not to mention the Rybka's disassembling and copying of the tables he said.

[Andrej Sidorov]

In programming 'idea' and 'algorythm' are very close. I would say, the same things. Alexander Shmidt told to Rajlich:

We had our first contact when I had questions about a similarity to Fruit in the search, others found similarities in the evaluation. Some people where a little bit suspicious that Rybka could be a clone of the open source engine.

Why shouldn’t Rajlich discover code of Rybka to experts (as Osipov did) to prove that he didn't copy parts of Fruit? Or didn't rewrite Fruit as Osipov claims?

[Onno Garms]

From this and from the important "This operation had not changed anything" means he has copied Fruit's source and rewritten it to use bitboards.

How can you "copy and rewrite"? Either you copy and modify, or you rewrite. As Dan Corbit did not tell anything about code stolen from Fruit, obviously the latter has been done. Moreover, even without knowing Russian language, isn't it clear from Dan's expertise that we should not take "not changed anything" too literally?

[Dann Corbit]

Strelka is not written in the style of fruit.

Fruit consists of three times as many C++ files as the C files of Strelka.
Now that I know the heritage, I can spot (for instance) that pawn_info_t of strelka and fruit are very similar. There are twice as many structure types in Fruit as there are in Strelka. There are 55 functions in Strelka and 257 functions in Fruit 22. Strelka is a bitboard engine and fruit is not. The style of Strelka's author is not like Fabian (does not use the assert() masterfully like Fabian for instance -- not a single assert in the whole code base).

The code is quite significantly rewritten, but I can now see the Fruit skeleton for some functions.