Belka,Glaurung 2.01 smp, Rybka, DS11,Zappa on Xeon

[GenoM]

you're asking me to walk this circle again... What kind of evidence is published in this thread? A statement of Rajlich that Strelka is Rybka 1.0 clone? Come on, we're not kids...

[Tony Thomas]

I think there are some strong arguments here

http://rybkaforum.net/cgi-bin/rybkaforu ... =2691;pg=1

Including the finding that strelka is more similar to rybka beta than rybka beta is later rybkas. there is other evidence presented in this thread that i find complelling. Too many coincidences. of course if someone can refute this evidence in some way i would be interested

Osipov admitted to taking the table from Rybka, it could be the table that's causing all those weird coincidences.

[Uri Blass]

you're asking me to walk this circle again... What kind of evidence is published in this thread? A statement of Rajlich that Strelka is Rybka 1.0 clone? Come on, we're not kids...

I think that it is clear that strelka is based on rybka and we do not need the code to know it.

The output is enough to know that strelka was not developed in order to be stronger engine but in order to be similiar to rybka.

Strelka even share the same strupid bugs that only rybka has.
You can see that the evaluation drops from -3 to -7 and rybka beta has the same behaviour when no other program will show that type of behaviour.


New game - Rybka 2.3.2a 32-bit
[d]rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKB1R w KQkq - 0 1

Analysis by Strelka 1.8 UCI:

1.e2-e3
-+ (-2.66) Depth: 1 00:00:00
1.d2-d3 Nb8-c6
-+ (-3.03) Depth: 2 00:00:00
1.d2-d3 Nb8-c6 2.Bc1-f4
-+ (-2.84) Depth: 3 00:00:00
1.d2-d4 Nb8-c6 2.d4-d5 Nc6-e5
-+ (-3.21) Depth: 4 00:00:00
1.d2-d4 Ng8-f6 2.Bc1-f4 Nb8-c6 3.Qd1-d3
-+ (-3.14) Depth: 5 00:00:00
1.d2-d4 Ng8-f6 2.d4-d5 e7-e6 3.Bc1-g5 Bf8-c5
-+ (-3.25) Depth: 6 00:00:00 6kN
1.d2-d4 Ng8-f6 2.d4-d5 d7-d6 3.Bc1-f4 Nf6-g4 4.Qd1-d4
-+ (-7.30) Depth: 7 00:00:00 31kN
1.d2-d4 e7-e6 2.Bc1-f4 c7-c5 3.d4xc5 Qd8-a5+ 4.Bf4-d2 Qa5xc5 5.c2-c3
-+ (-7.25) Depth: 8 00:00:00 555kN
1.d2-d4 e7-e6 2.d4-d5 e6xd5 3.Qd1xd5 Bf8-e7 4.Qd5-e4 Ng8-f6 5.Qe4-d3
-+ (-7.25) Depth: 9 00:00:02 926kN

(, 30.11.2007)

New game - Rybka 2.3.2a 32-bit
rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKB1R w KQkq - 0 1

Analysis by Rybka 1.0 Beta 32-bit:

1.d2-d4
-+ (-3.14) Depth: 3 00:00:00
1.d2-d4
-+ (-3.23) Depth: 4 00:00:00
1.d2-d4 Ng8-f6
-+ (-7.30) Depth: 5 00:00:00
1.d2-d4 c7-c6 2.Bc1-f4
-+ (-7.30) Depth: 6 00:00:00 120kN
1.d2-d4 c7-c6 2.Bc1-f4 Ng8-f6
-+ (-7.30) Depth: 7 00:00:00 190kN
1.d2-d4 c7-c6 2.Bc1-f4 Qd8-b6 3.Bf4-c1
-+ (-7.30) Depth: 8 00:00:01 702kN

(, 30.11.2007)

Note that Vaik Knows that it is a bug and he fixed this behaviour in rybka1.1 and later versions

New game - Rybka 2.3.2a 32-bit
rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/R1BQKB1R w KQkq - 0 1

Analysis by Rybka 1.1 32-bit:

1.d2-d4
-+ (-3.51) Depth: 2 00:00:00
1.d2-d4
-+ (-3.58) Depth: 3 00:00:00
1.d2-d4 Ng8-f6
-+ (-4.14) Depth: 4 00:00:00
1.b2-b3 Nb8-c6
-+ (-4.10) Depth: 4 00:00:00
1.b2-b3 Nb8-c6 2.Bc1-b2
-+ (-4.03) Depth: 5 00:00:00
1.b2-b3 Nb8-c6 2.Bc1-b2 Ng8-f6
-+ (-4.10) Depth: 6 00:00:00 154kN
1.b2-b3 Nb8-c6 2.Bc1-b2 Ng8-f6 3.a2-a3
-+ (-4.03) Depth: 7 00:00:04 1448kN
1.b2-b3 Nb8-c6 2.Bc1-b2 Ng8-f6 3.a2-a3 d7-d6
-+ (-4.10) Depth: 8 00:00:05 1870kN
1.b2-b3 e7-e6 2.e2-e3 b7-b6 3.g2-g4 Bc8-b7 4.Rh1-g1
-+ (-4.07) Depth: 9 00:00:42 14890kN

(, 30.11.2007)

[ozziejoe]

I think there are some good arguments, especially made by turbojuice.

here are some arguments, with my thoughts with "****" by them. the original source of these quotes can be found in http://rybkaforum.net/cgi-bin/rybkaforu ... =2691;pg=1

"""""
A major mistake made by the many knowledgeable and respected people on CCC at the time of the Strelka talk was that it was nearly impossible to reverse engineer the Rybka code--this was one of the main working assumptions involved in the initial decision that Strelka might not be a clone. However, this assumption has been shown to be false.

A second major mistake by the many knowledgeable and respected people on CCC at the time of the Strelka talk was one of basic logic: some of them said that by looking at the Strelka code, they could tell that it was different from Rybka's code. But they never saw Rybka's code! This they cannot say that the code wasn't copied from Rybka.

""""

********Here is something from vas..

""""""
at the top of the page). Here are some examples from Vas:

At this point we don't know how Strelka was made. If I had to bet, I'd put my money on straight decompile with changes.

Vas

Also from http://rybkaforum.net/cgi-bin/rybkaforu ... s#pid18939 at a different point. I should add that the explanation above that by the person calling himself "Yuri Osipov" is absolute garbage: the probability of being able to match Rybka weighings with a host of different positions is very small to begin with if the selection is more than just a few. The probability that the evaluations of new positions would then still match is so incredibly small that one would have to be extremely foolish to believe the claim. Anyway, from that thread:

Note that this explanation is itself bogus:

a) 'Osipov' claims that he changed the Fruit board representation from mailbox to bitboard and got a 2x speedup in performance. This is simply a clueless comment, there would be no speedup of anywhere near this magnitude.
b) 'Osipov' claims to take only Rybka's eval and search, yet Strelka 1.8 uses Rybka's exact UCI output strings.
c) 'Osipov' claims that he added a Winboard parser (after date of Fruit 2.1 release) so that Strelka could play in Ridderkerk - another clueless comment.
d) ...

I will think a little bit about this. Maybe what I should do is get the Strelka 1.8 source code, claim it as my own, and release it under GPL.

With these anonymous cloners who risk nothing, and with current Rybka level, computer chess may be headed for some turbulence.

Vas
*********


*******and finally, i find the following from turbojuice compelling (and have seen the data on this)

"
I should note that one cannot underestimate the importance of the fact that in a decent percentage of positions, not only does Strelka match Rybka's eval exactly, but the evals at successive depths and times are also matched exactly. The only difference is that Strelka doesn't "correctly" count the nodes calculated--it does it more in the "old Fritz" style instead of the "new Rybka" style, and that's if it even does it in a correct way at all. These examples have been analyzed many times in the various fora and repeatedly confirmed by others. I don't think that it's really up for debate anymore that Strelka is a Rybka clone: it's very clear that it is. "


****


*****I think this evidence is close to convincing. I also find alot of uri's arguments compelling (similarity in evaluation by depth).
I would be happy to here counter-arguments, refutations of this evidence. Here are some arguments that i do not find convincing

Unconvincing argument 1:

1. "it is just your opinion that strelka is a clone. I have my opionion." I could be of the opinion that the world is flat. Saying that a cliam is an "opinion" does nto exempt it from needing evidence

Unconvincing argument2:

2. "You guys are just worshiping at the feet of vas" or related arguments. This is ad hominum. It is meaningless in the relm of evidence.


I am open to the possibility that there is a reasonable explanations for all this stuff. Maybe strelka is not a clone, but i just don't understand thse similarities. Ideally, you would maybe generate 100 representative positions and than measure similairty in evaluation. You could do this with enough engines that you would have a baserate level of similarity. then you could conduct statistical tests (e.g., you could ask, "is the similarity between rybka and strelka much higher than would be expected based on similarity of other engines"

As you can see, I am not convinced either way but am leaning towards the clone argument

[George Tsavdaris]

Just a hypothesis: he (Rajlich) was affraid to have any relationships with a man (Osipov) whos intentions he (Rajlich) can hardly understand.

And what are his intentions?

[GenoM]

Just a hypothesis: he (Rajlich) was affraid to have any relationships with a man (Osipov) whos intentions he (Rajlich) can hardly understand.

And what are his intentions?

Why asking me?! Ask Osipov. May be he wanted (or hoped for) that if Rajlich look at the sources of Strelka he'll see that Strelka isn't a clone of Rybka. Who knows?

[Alessandro Scotti]

I think there are some good arguments, especially made by turbojuice.

None of the above comes even close to prove that Strelka is a clone. The only explicit hint ("straight decompile with changes") has been already disproved by the many people who have received clearly formatted and commented C source code for Strelka, which with current technology is impossible to achieve.

Instead, Strelka's author explicitly claimed that he reverse engineered parts of Rybka in order to have Strelka behave as close as possible to Rybka. So yes, there are many examples where Strelka and Rybka have same evaluation: this is by design.

But, reverse engineering is not cloning, far from it! Reverse engineering has been a powerful tool for faster progress in computing and better competition on the market, and as such is offered protection by law in most countries. (For example a free operating system like Linux is only possible because of the many private and closed source protocols, device drivers, services, etc. that have been reverse engineered in order to obtain a corresponding free implementation).

I think it's better to avoid terms like "clone" or "illegal" at this time.