Hustle 5.01 Uci Win32 available(fixed some bugs)

[playjunior]

So Pedro cloning of Strelka is allowed?
And who is the expert that judged Strelka is a clone? Where is the evidence? "Vasik told so" isn't valid

Regards,
Geno

Have you seen the threads in Rybka forum?
Some guys disassembled Strelka 1.0 or 1.8, i don't remember which one exactly, and posted results in a thread.

There were:
1)Huge segments of data with same values. Means-Rybka eval and material imbalance tables were copied.
2)Segments of code that corresponded to the search. Was the same. Two different search functions do not produce the same assembly code. Even if you give a same algorithm and they have 2 (equivalent!) implementations, they are very unlikely to produce the same assembly code. If the algorithm is complex enough. Rybka's search is complex enough, so we have to deal with something extraordinary here.

Then, there were a couple of threads where people posted positions were Rybka had faulty evaluation/bug, and Strelka too. They tried other engines -none produced the bug. It's not the same eval, the same line-its a BUG.


Some people claim everything is a clone. For them-evidence is a blind belief in good Vasik and evil russians. When they see same UCI strings and similar playing style, it's over. You cannot change their mind.

In the begining Vas was not saying Strelka is a clone. He was saying that, say, displaying +-320 as a mate score is very natural (he said it's something like the biggest value you can store in a hash entry), not implementing underpromotion is very natural too, and people should not make claims about clone based on that. However, after they posted the search stuff and the data tables, he pretty much changed his opinion.

I don't know what is enough evidence for you. For me, however, same assembler footprint for the SEARCH, same CONSTANTS, same BUGS is more than enough.

[geots]

So Pedro cloning of Strelka is allowed?
And who is the expert that judged Strelka is a clone? Where is the evidence? "Vasik told so" isn't valid

Regards,
Geno

Have you seen the threads in Rybka forum?
Some guys disassembled Strelka 1.0 or 1.8, i don't remember which one exactly, and posted results in a thread.

There were:
1)Huge segments of data with same values. Means-Rybka eval and material imbalance tables were copied.
2)Segments of code that corresponded to the search. Was the same. Two different search functions do not produce the same assembly code. Even if you give a same algorithm and they have 2 (equivalent!) implementations, they are very unlikely to produce the same assembly code. If the algorithm is complex enough. Rybka's search is complex enough, so we have to deal with something extraordinary here.

Then, there were a couple of threads where people posted positions were Rybka had faulty evaluation/bug, and Strelka too. They tried other engines -none produced the bug. It's not the same eval, the same line-its a BUG.


Some people claim everything is a clone. For them-evidence is a blind belief in good Vasik and evil russians. When they see same UCI strings and similar playing style, it's over. You cannot change their mind.

In the begining Vas was not saying Strelka is a clone. He was saying that, say, displaying +-320 as a mate score is very natural (he said it's something like the biggest value you can store in a hash entry), not implementing underpromotion is very natural too, and people should not make claims about clone based on that. However, after they posted the search stuff and the data tables, he pretty much changed his opinion.

I don't know what is enough evidence for you. For me, however, same assembler footprint for the SEARCH, same CONSTANTS, same BUGS is more than enough.

Must we suffer thru all this Strelka-Belka crap again. Dear God.

[GenoM]

Must we suffer thru all this Strelka-Belka crap again. Dear God.

Yes, George, you'll suffer it again and again and again
Accept your fate

[GenoM]

Hi Andranik
If Strelka is clone you have to have a version of how Osipov got access to the Vasik's code.
If your answer is disassembling, would you say that the code of Strelka was got disassembling way.
And third question: which part of software has to be exactly copied to name some program clone? Or how many percent of code of other programe it has to have to be a clone for you?
Thanks.

[playjunior]

Vasik told he has a dedicated computer with Rybka source on it, and it is not connected to internet. So-disassembling seems the only way.

If you have have a brief look at Strelka's source, you will see an unprecedentedly high number of hard-coded constants (a sure sign of poor technique or disassembly). Indeed, this is a much, much less convincing argument compared to same assembly code for the search function.

The third question: what is a clone?-reminds me of an episode from the Simpsons, where Bart Simpson has to deal with a moral dilemma. He is "employed" by some Italian mafia guys, and he finds out that they sell cigarettes illegally. He asks "but isn't that stealing?". Then goes a nice piece of reasoning:
Mafia Boss: "Look, Bart boy, is it stealing if you take a little bread to feed your starving family?"
Bart: "Nooo."
M. B. "And if your family is big? And you take some bread, not really a little?"
Bart(confused): "Not, I guess?"
M.B.: "Correct! And if you do not need only to feed them, but buy clothes and so on? And you do not take bread, but you just sell cigarettes without telling the government, and buy break and little other things on that money!"

We are in a similar situation here. I don't care what is a concrete and precise definition of a clone. Sometimes everything is clear enough without a mathematical definition.

As I said before, there is no way to get same assembly code for the brain of a program without copying. There is no way to have the same constants without copying. You never get exactly same bugs. It never happens. This is something you have to deal with...

[Graham Banks]

Links to illegal clones will not be permitted on CCC.

(the three moderators)