The rumors about Strelka beeing a Rybka clone is still alive

[Alexander Schmidt]

I don't understand why this should prove that Strelka is a clone. This is a primitive position, can it not be that two or more engines evaluate it the same?
In positions with more material I cannot see any similarities in their evaluation output.

I still had doubts with the positions shown by others with nearly the same eval in some analysis line, or with the same eval in a few lines. But with Kvs.K positions you have always the same eval in every analysis line in all positions.

This is enough for me to be absolutely sure Strelka has parts of Rybka inside.

Best,
Alex

[Roman Hartmann]

But, you ignore the possibility of what Tord has written!

The possibility of a 6328 byte block being completely identical??? LOL
You don´t seem to know much about probability ...

And BTW, it can´t have to do anything with opening books, because both engines don´t support it!

Why has Strelka to be the clone? Rybka and Strelka have some things in common, that's pretty obvious.

Maybe Strelka and Rybka are both based partly on some other third party code/work which was never published. At least they make use of some new ideas which no one beside them is using.
Strelka and Rybka might just share the same roots (and therefore also some code).

Maybe Vas could bring some light into that matter.

Roman

[James Constance]

Yes this is a possibility. Also, if they share code, Rybka could be a clone of Strelka.

[Christopher Conkie]

I don't understand why this should prove that Strelka is a clone. This is a primitive position, can it not be that two or more engines evaluate it the same?
In positions with more material I cannot see any similarities in their evaluation output.

I still had doubts with the positions shown by others with nearly the same eval in some analysis line, or with the same eval in a few lines. But with Kvs.K positions you have always the same eval in every analysis line in all positions.

This is enough for me to be absolutely sure Strelka has parts of Rybka inside.

Best,
Alex

Weights

Could you send me a mail please Alex. I have something I want you to see.

I don't have your mail address to hand right now you see.

Regards

Christopher

[George Tsavdaris]

Can´t you imagine that it might be just a small typo when writing such hex addresses?
In fact the Rybka address range should be 25AC8-2737F hex, and this makes both blocks the same size.

Well that makes sense too....

(BTW, 6328 bytes - that for your ´basic logic´ ... )

Yes absolutely.
I just wanted to compare the bytes of the 2 programs for the given range of addresses so the +1 would not make any difference.
But since i wrote that the bytes of Strelka was ...blabla, then you are right to correct me since the actual bytes were +1 of what i have given....

They are not similar, the are not almost identical - they are absolutely identical!
Coincidence?

No way i guess!

Too bad for Strelka....

[George Tsavdaris]

Both Rybka 1.0 and Strelka don't contain any EGTB code. Though I agree it could be some library code, theoretically.

It could be some library code and coincidentally only Rybka 1.0 beta and Strelka have it? No other program.
OK we may accept this.

What about the positions Uri posted? This has nothing to do with having the library but only on program's evaluation and search. Rybka's X ply on many positions is exactly the same with Strelka's in X+2 ply for many different plies.
And the X/X+2 pattern is there on many positions. That would be enough by itself.
Coincidence again? I don't believe in double coincidences....

And what about K-K positions? Only Rybka and Strelka has this behavior.
Coincidence again?

Well this is too much.....

[Uri Blass]

I don't understand why this should prove that Strelka is a clone. This is a primitive position, can it not be that two or more engines evaluate it the same?
In positions with more material I cannot see any similarities in their evaluation output.

I still had doubts with the positions shown by others with nearly the same eval in some analysis line, or with the same eval in a few lines. But with Kvs.K positions you have always the same eval in every analysis line in all positions.

This is enough for me to be absolutely sure Strelka has parts of Rybka inside.

Best,
Alex

Here is more evidence for similiarity.

It seems strange that both rybka and strelka has evaluation drop from +15.xx to +10.xx in KQ vs K so it seems that both have the same bug

It is not logical that 2 different programmers will have the same bugs by an accident.

New game
[d]4k3/8/8/8/8/8/8/3QK3 w - - 0 1

Analysis by Rybka 1.0 Beta 32-bit:

1.Qd6
+- (14.99) Depth: 3 00:00:00
1.Qd6
+- (15.12) Depth: 4 00:00:00
1.Qd6 Kf7
+- (10.18) Depth: 5 00:00:00
1.Qd6 Kf7 2.Kd2
+- (10.18) Depth: 6 00:00:00 34kN
1.Qd6 Kf7 2.Kd2 Kg7
+- (10.18) Depth: 7 00:00:00 57kN
1.Qd6 Kf7 2.Kd2 Kg7 3.Qe6
+- (10.18) Depth: 8 00:00:00 94kN
1.Qd6 Kf7 2.Kd2 Kg7 3.Qe6 Kf8
+- (10.18) Depth: 9 00:00:00 150kN
1.Qd6 Kf7 2.Kd2 Kg7 3.Qe6 Kf8 4.Qf6+ Ke8
+- (10.18) Depth: 10 00:00:00 227kN
1.Qd6 Kf7 2.Kd2 Kg7 3.Qe6 Kf8 4.Qf6+ Ke8 5.Qe5+ Kf7
+- (10.18) Depth: 11 00:00:01 345kN
1.Qd6 Kf7 2.Kd2 Kg7 3.Qe7+ Kh8 4.Ke3 Kg8 5.Kf4
+- (320.00) Depth: 12 00:00:01 495kN

(Uri, MyTown 09.07.2007)

New game
4k3/8/8/8/8/8/8/3QK3 w - - 0 1

Analysis by Strelka 1.8 UCI:

1.Qd5
+- (15.34) Depth: 1 00:00:00
1.Qd5 Ke7 2.Ke2
+- (15.34) Depth: 3 00:00:00
1.Qd6 Kf7 2.Ke2 Kg7
+- (15.48) Depth: 4 00:00:00
1.Qd6 Kf7 2.Ke2 Kg7 3.Ke3
+- (15.56) Depth: 5 00:00:00
1.Qd6 Kf7 2.Ke2 Kg7 3.Ke3
+- (15.56) Depth: 6 00:00:00
1.Qd6 Kf7 2.Kd2 Kg7 3.Qe6 Kf8
+- (15.69) Depth: 6 00:00:00 30kN
1.Qd6 Kf7 2.Kd2 Kg7 3.Qe6 Kf8 4.Qf6+ Ke8
+- (10.59) Depth: 7 00:00:00 131kN
1.Qd6 Kf7 2.Kd2 Kg7 3.Qe6 Kf8 4.Qf6+ Ke8 5.Qe5+ Kf7
+- (10.59) Depth: 8 00:00:00 294kN
1.Qd6 Kf7 2.Kd2 Kg7 3.Qe6 Kf8 4.Qf6+ Ke8 5.Qe5+ Kf7 6.Qf4+ Ke7
+- (10.59) Depth: 9 00:00:00 403kN
1.Qd6 Kf7 2.Kd2 Kg7 3.Qe6 Kf8 4.Qf6+ Ke8 5.Qe5+ Kf7 6.Qf4+ Ke7 7.Qg5+ Ke6
+- (10.59) Depth: 10 00:00:00 573kN
1.Qd6 Kf7 2.Kd2 Kg7 3.Qe6 Kf8 4.Qf6+ Ke8 5.Qe5+ Kf7 6.Qf4+ Ke7 7.Qg5+ Ke6 8.Qe3+ Kd6
+- (10.59) Depth: 11 00:00:00 818kN
1.Qd6 Kf7 2.Kd2 Kg7 3.Qe6 Kf8 4.Qf6+ Ke8 5.Qe5+ Kf7 6.Qf4+ Ke7 7.Qg5+ Ke6 8.Qe3+ Kd6 9.Qb6+ Ke7
+- (10.59) Depth: 12 00:00:00 1161kN
1.Qd6 Kf7 2.Kd2 Kg7 3.Qe6 Kf8 4.Qf6+ Ke8 5.Qe5+ Kf7 6.Qf4+ Ke7 7.Qg5+ Ke6 8.Qe3+ Kd6 9.Qb6+ Ke7 10.Qb4+ Ke6
+- (10.59) Depth: 13 00:00:01 1632kN
1.Kd2 Ke7 2.Kd3 Ke6 3.Ke4 Kf6 4.Qd5 Ke7 5.Kf5 Kf8 6.Qd7 Kg8 7.Kf6 Kf8 8.Qe7+
+- (#8) Depth: 14 00:00:01 2736kN

(Uri, MyTown 09.07.2007)

[Dr.Wael Deeb]

Sorry!

When single quotes are put around a word it means not to take the litteral meaning but to understand the spirit of what it is intended to convey.

I suggested an alternative explaination and was laughed at and told that I had a poor understanding of things. Then I was accused of wildcatting. This is just wrong. But, I am done with it.

The only further point that I would like to make, is that I hope my mom does not see this. I do not want her to think that I swore or something!

Hi Michael,
Everybody knows that you are an excellent programmer and have a creatve soul keeping on going on and on toward perfiction,so don't get offended
As for Franz,i will pull his ears whenever i put my hands on him

[F.Huber]

As for Franz,i will pull his ears whenever i put my hands on him

Oh, then one thing is clear for me: I´ll never make holidays in Bulgaria!

[Alexander Schmidt]

It seems strange that both rybka and strelka has evaluation drop from +15.xx to +10.xx in KQ vs K so it seems that both have the same bug

It is not logical that 2 different programmers will have the same bugs by an accident.

Hi Uri,

the problem is, some of such positions are no real evidence. At least for me. Of course there where many of this positions, especially with v1.8 of Strelka. I didn't look at all of the positions of the first Strelka release, but what I saw wasn't a evidence at all because I always found other engines with the same similaries as the ones posted.

To say Engine X is a clone or has the same roots, you need exactly the same behaviour in a way.

Best,
Alex