CCRL - Supports clones and illegal engines ?

[Michael Sherwin]

Hello Daniel. I am very disappointed by your attitude, by seeing that you think it is acceptable to make such demands.

I can see why Daniel feels annoyed, but when the experts can't agree over whether Strelka is a clone, or whether it's illegal, why should we be expected to make such an assumption?
Under such circumstances, it's also unfair to be accused of a lack of ethics or morals.
If Strelka was proven to be illegal in a court of law, there would be no questions. We'd dump it immediately.

Regards, Graham.

Assuming that we have heard the truth about how Strelka was done.

1) Strelka started off as a rewrite of Fruit into bitboards. Fruit has almost no original ideas in it and the translation to bitboards requires a large change to the data structure. Strelka is therefore not a Fruit clone and not in violation of the GPL.

2) Strelka 1.0 is suspect due to copying of the data tables directly from Rybka. This is a grey legal area, because, in general pure data is not copyrightable. The format of the data may be though. However, Strelka 1.8 solves this by creating functionals that reproduce the same end result that the data was used for. This is a true advancement to computer chess. Strelka 1.8 is not a clone based on point 2.

3) Strelka was given other eval parameters that cause it to evaluate very similar to Rybka. Plus the search output of Strelka is a lot more 'honest'. If tunning of Strelka led to the same evaluations as Rybka then the matter would be decided. Strelka 1.8 would not be a clone or illegal in any sense that matters. However, the person claiming to be the author of Strelka indicated that he did not tune Stelka to be like Rybka, but rather, took the parameters directly from Rybka in order to make Stelka play like Rybka. This last point then does indicate that Strelka has cloned the eval of Rybka. The eval is most of a programs identity. This IMO does make Srelka a Rybka clone.

[Graham Banks]

Hello Daniel. I am very disappointed by your attitude, by seeing that you think it is acceptable to make such demands.

I can see why Daniel feels annoyed, but when the experts can't agree over whether Strelka is a clone, or whether it's illegal, why should we be expected to make such an assumption?
Under such circumstances, it's also unfair to be accused of a lack of ethics or morals.
If Strelka was proven to be illegal in a court of law, there would be no questions. We'd dump it immediately.

Regards, Graham.

Assuming that we have heard the truth about how Strelka was done.

1) Strelka started off as a rewrite of Fruit into bitboards. Fruit has almost no original ideas in it and the translation to bitboards requires a large change to the data structure. Strelka is therefore not a Fruit clone and not in violation of the GPL.

2) Strelka 1.0 is suspect due to copying of the data tables directly from Rybka. This is a grey legal area, because, in general pure data is not copyrightable. The format of the data may be though. However, Strelka 1.8 solves this by creating functionals that reproduce the same end result that the data was used for. This is a true advancement to computer chess. Strelka 1.8 is not a clone based on point 2.

3) Strelka was given other eval parameters that cause it to evaluate very similar to Rybka. Plus the search output of Strelka is a lot more 'honest'. If tunning of Strelka led to the same evaluations as Rybka then the matter would be decided. Strelka 1.8 would not be a clone or illegal in any sense that matters. However, the person claiming to be the author of Strelka indicated that he did not tune Stelka to be like Rybka, but rather, took the parameters directly from Rybka in order to make Stelka play like Rybka. This last point then does indicate that Strelka has cloned the eval of Rybka. The eval is most of a programs identity. This IMO does make Srelka a Rybka clone.

Hi Michael,

just as a matter of interest, what's your view on Toga?

Regards, Graham.

[Uri Blass]

The threshold of what is legal or illegal will never be agreed upon. Strelka is definetely not kosher but I guess many people were happy that an engine was able to perform well out of the blue. This is really not fair to engine developers that spent a lot of time and effort creating original engines. Why spend all the effort on making an original work when you can simply copy or clone an already strong engine. The UCI strings of Strelka are identical to Rybka 1.0 according to Vas, if this was done on purpose (for whatever idiotic reason) to try to mimick Rybka then I think the author should be treated as a cloner. If he wants to get away with cheating then he should do a good job at that and try to mask his effort a little better. As it is now ... accepting Strelka as an original work is a farse.

Your assumption that it is simple to copy or clone already strong engine is not proved and it is possible that it took more time to write strelka relative to the time that vasik needed to write rybka.


Your assumption that the author wanted to get away with cheating is not correct and it is clear that his target was different.

I guess that he simply wanted people to know the information that rybka was hiding about the first plies.

Uri

[Uri Blass]

Daniel,

Don't you have more important things to do than to spend time and energy to discuss decisions made internally by CCRL?

I can tell you that the "Strelka" thread in our discussion board isn't small, by no means.

As for this thread, you could have kept it as a private message/email with Graham, no need to publish this.

Further, if you are unhappy with CCRL's current stance on the Strelka matter, you could go public stating that you will make Homer private and such preventing us from testing it if that is what you really want.

As it stands now, Homer is a free engine and we in CCRL enjoy testing it.


Think about this scenario:

Is it possible that Yuri Osipov invented the wheel independently from Vasik Rajlich?

I am sure you get my analogy.

Regards,
Martin

If the wheel is complicated enough then it is clear that it is practically impossible that the sides invented it independently.

By the same logic you can claim that maybe 2 people had the same ideas and wrote exactly the same book.

In theory it is possible but practically the court should decide that it is impossible.

both strelka and rybka have the same bugs that other programs do not have:

1)same evaluation that intially is not 0.00 in king againt king.

2)drop in score relative to first plies when one side has big material advantage.

New game - Rybka 2.3.2a 32-bit
[d]1nb1kbn1/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQ - 0 1

Analysis by Rybka 1.0 Beta 32-bit:

1.Nb1-c3
+- (19.63) Depth: 3 00:00:00
1.Nb1-c3
+- (19.32) Depth: 4 00:00:00
1.d2-d4
+- (19.46) Depth: 4 00:00:00
1.d2-d4 Nb8-c6
+- (18.42) Depth: 5 00:00:00
1.d2-d4 Nb8-c6 2.d4-d5
+- (18.48) Depth: 6 00:00:00 135kN
1.d2-d4 Nb8-c6 2.d4-d5 Nc6-b4
+- (18.42) Depth: 7 00:00:00 294kN
1.d2-d4 Nb8-c6 2.d4-d5 Nc6-b4 3.c2-c3
+- (18.48) Depth: 8 00:00:03 1733kN
1.d2-d4 c7-c6 2.d4-d5 c6-c5 3.c2-c3 a7-a6
+- (18.48) Depth: 9 00:00:04 2161kN
1.d2-d4 c7-c6 2.d4-d5 c6-c5 3.c2-c3 a7-a6 4.Bc1-f4
+- (18.48) Depth: 10 00:00:07 3395kN
1.d2-d4 c7-c6 2.d4-d5 c6-c5 3.c2-c3 Ke8-d8 4.Bc1-f4 d7-d6
+- (18.48) Depth: 11 00:00:09 4242kN
1.d2-d4 c7-c6 2.d4-d5 c6-c5 3.c2-c3 Ke8-d8 4.Bc1-f4 d7-d6 5.Qd1-a4
+- (18.48) Depth: 12 00:00:14 6148kN

(, 19.07.2007)


New game - Rybka 2.3.2a 32-bit
1nb1kbn1/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQ - 0 1

Analysis by Strelka 1.8 UCI:

1.Ng1-f3
+- (19.99) Depth: 1 00:00:00
1.Ng1-f3 Nb8-c6
+- (19.33) Depth: 2 00:00:00
1.Ng1-f3 Nb8-c6 2.Nb1-c3
+- (20.00) Depth: 3 00:00:00
1.Ng1-f3 Nb8-c6 2.Nb1-c3 Ng8-f6
+- (19.36) Depth: 4 00:00:00
1.Ng1-f3 Nb8-c6 2.Nb1-c3 Ng8-f6 3.d2-d4
+- (19.63) Depth: 5 00:00:00
1.d2-d4 Nb8-c6 2.d4-d5 Nc6-b4 3.c2-c3 Nb4-a6
+- (19.46) Depth: 6 00:00:00 10kN
1.d2-d4 Nb8-c6 2.d4-d5 Nc6-b4 3.c2-c3 Nb4-a6 4.Bc1-g5
+- (18.42) Depth: 7 00:00:00 31kN
1.d2-d4 Nb8-c6 2.d4-d5 Nc6-b4 3.c2-c3 Nb4-a6 4.Bc1-g5 d7-d6
+- (18.48) Depth: 8 00:00:00 491kN
1.d2-d4 Nb8-c6 2.Bc1-f4 d7-d6 3.d4-d5 Nc6-b4 4.c2-c3 Nb4-a6 5.Qd1-a4+ Ke8-d8
+- (18.48) Depth: 9 00:00:00 1041kN
1.d2-d4 Nb8-c6 2.Bc1-f4 d7-d6 3.d4-d5 Nc6-b4 4.c2-c3 Nb4-a6 5.e2-e3 Ke8-d8
+- (18.48) Depth: 10 00:00:01 1890kN
1.d2-d4 Nb8-c6 2.Bc1-f4 d7-d6 3.d4-d5 Nc6-b4 4.c2-c3 Nb4-a6 5.Qd1-d3 Ng8-f6 6.Qd3-b5+ Ke8-d8
+- (18.48) Depth: 11 00:00:02 2299kN
1.d2-d4 Nb8-c6 2.Bc1-f4 d7-d6 3.d4-d5 Nc6-b4 4.c2-c3 Nb4-a6 5.Qd1-d3 Ng8-f6 6.e2-e3 Ke8-d8
+- (18.48) Depth: 12 00:00:04 5700kN
1.d2-d4 a7-a6 2.Nb1-c3 Ng8-f6 3.Ng1-f3 d7-d6 4.Ra1-b1 Nf6-g8 5.Nf3-g5 Bc8-f5 6.Nc3-d5 c7-c6 7.Nd5-c7+ Ke8-d8
+- (18.48) Depth: 13 00:00:06 9021kN
1.d2-d4 a7-a6 2.Nb1-c3 Ng8-f6 3.Ng1-f3 d7-d6 4.Ra1-b1 Nf6-g8 5.Nf3-g5 Bc8-f5 6.Nc3-d5 c7-c6 7.Nd5-c3 Ng8-f6
+- (18.48) Depth: 14 00:00:12 19304kN
1.d2-d4 a7-a6 2.Nb1-c3 d7-d6 3.Qd1-d3 g7-g6 4.Qd3-c4 c7-c6 5.d4-d5 c6-c5 6.Qc4-a4+ Nb8-d7 7.Qa4-a5 b7-b6 8.Qa5-a4 h7-h6
+- (18.48) Depth: 15 00:00:34 56457kN

(, 19.07.2007)

[Daniel Mehrmann]

Hello Daniel. I am very disappointed by your attitude, by seeing that you think it is acceptable to make such demands. I personally did not test Homer yet, and reading this thread I think there is very little chance I'll ever test it.

Homer is a free chess engine, and as such it is a part of complex computer chess picture. People interested in computer chess have right to know where Homer stands in that picture. Our goal is to provide such information to anyone interested in computer chess. If you don't want us (or anyone) to test future version of Homer, I suggest you to make it private, thus removing it from the picture of free chess engines.

Well, you're smart and tricky. You're starting slightly to change the topic and go away of Strelka. You give me no proofs that i could say i'm wrong at all. You're not really interested in Homer, that's no problem for me of course, your interest seems big ELO points and the "price" doesn't matter.

Regarding Strelka, I personally agree that it plays very similar to Rybka 1.0. However I have no idea whether Strelka is legal or not. The GPL violation was not confirmed by the experts who saw Strelka code. I am also not sure on what you are basing your claims that Strelka contains stolen code from Rybka.

You have no idea if its legal or not ? You're not sure if code is stolen from Rybka ? You're looking only to the experts ?



I'm sorry, but this answer is poorless. It seems to me you're hiding behind the "experts" and you aren't telling me your own view about these points.
You given me no new proofs that i could be wrong. Well, i think you lost the overview at all and you should first think about yourself to build your own opinion. Maybe you should think about what will happend in the future of computerchess if you allow such engines now. I believe it's the beginning of the end if we follow this way...

We are trying to provide objective comparison of chess engines. If you don't want to compete with Strelka (or any other engines stronger than yours), you can run your own tournaments and publish the results. You are free to do that. But you better stop harrasing the testers who invest so much of their time and efforts into doing this free service to the community.

Ah, again you're trying to change the topic at all and you're trying that i'm looking bad now in this talk, because i would do a "harrasing attack" and wouldn't accept the work of CCRL.
Nice try, but i'm sure it doesn't work. Well, yes i'm trying to support a provide objective comparison between the engines. Thats a goal, so i'm trying to keep Strelka out.

Kirill, you should ask yourself in a silent minute how much is your price for big ELO points ? Do you really wanna sells your moral view for 2750 points ?

Best,
Daniel

[Graham Banks]

Kirill, you should ask yourself in a silent minute how much is your price for big ELO points ? Do you really wanna sells your moral view for 2750 points ?

Best,
Daniel

Hi Daniel,

The strength of Strelka has nothing at all to do with the issue.
Likewise, how can what we're doing be morally wrong if we don't know for sure about Strelka being legal or otherwise.
Vas needs to say something. Until then, the axiom "innocent until proven guilty" should apply.

Regards, Graham.

[Daniel Mehrmann]

Kirill, you should ask yourself in a silent minute how much is your price for big ELO points ? Do you really wanna sells your moral view for 2750 points ?

Best,
Daniel

Hi Daniel,

The strength of Strelka has nothing at all to do with the issue.
Likewise, how can what we're doing be morally wrong if we don't know for sure about Strelka being legal or otherwise.
Vas needs to say something. Until then, the axiom "innocent until proven guilty" should apply.

Regards, Graham.

Graham, this won't work with Vas and you know that. You know no new proofs will come. No court decision will be done, it never happend in the whole history of computerchess.

I believe all these points are well known for you, but you're hiding behind it. By the way, I think Kirill can speak for himself or not ?

So, if i would say Homer is clone of engine X you wouldn't believe it so long another programmer verfiyed this option?

Again, I'm thinking i'm talking about the truth, because ELO points are the most importent factor for a ratinglist if engines going to be tested/listed or not.

Yes, we're talking about dark sides in computerchess now and for a long time it was a taboo anyway.

Best,
Daniel

[Graham Banks]

Again, I'm thinking i'm talking about the truth, because ELO points are the most importent factor for a ratinglist if engines going to be tested/listed or not.

Best,
Daniel

Have you seen the extensive list of amateur engines that we test?
I don't think you're being fair with that statement.

Regards, Graham.

[Daniel Mehrmann]

Again, I'm thinking i'm talking about the truth, because ELO points are the most importent factor for a ratinglist if engines going to be tested/listed or not.

Best,
Daniel

Have you seen the extensive list of amateur engines that we test?
I don't think you're being fair with that statement.

Regards, Graham.

Well Graham, that's a nice trick as well.
Of course many amateur are tested by the CCRL. But that's not the point. The point is that top engines are always interesting and basicly that's okay and i would prefer to test such engines as well. So, basicly there is a very high interest by the tester to test Strelka of course and that makes it difficult if problems comes up like in this case.

Best,
Daniel

BTW, I'm happy already that we have a open talk and that shows me that the CCRL is open for arguments.

[Graham Banks]

Again, I'm thinking i'm talking about the truth, because ELO points are the most importent factor for a ratinglist if engines going to be tested/listed or not.

Best,
Daniel

Have you seen the extensive list of amateur engines that we test?
I don't think you're being fair with that statement.

Regards, Graham.

Well Graham, that's a nice trick as well.
Of course many amateur are tested by the CCRL. But that's not the point. The point is that top engines are always interesting and basicly that's okay and i would prefer to test such engines as well. So, basicly there is a very high interest by the tester to test Strelka of course and that makes it difficult if problems comes up like in this case.

Best,
Daniel

BTW, I'm happy already that we have a open talk and that shows me that the CCRL is open for arguments.

Hi Daniel,

No tricks. Just telling it like it is.
Although many people are only interested in the top engines on big hardware, many people likewise have an interest in the amateurs and the weaker engines on 32-bit 1CPU.
I (and other testers) not only get enjoyment out of testing these engines, but we see it as a way of rewarding the efforts of their programmers.
One of our stated aims is get all of these engines 200+ games at 40/40 time controls.

Regards, Graham.