What's Strelka's Secret Sauce?

[bob]

Uri:

I understood that you were not claiming it was a fruit clone. I was being sarcastic in my post by claiming that since the eval was so close to Crafty's, it must be a crafty clone. This kind of "analysis" happens here all the time. Evaluations and/or PVs are simply not sufficient to prove clone/no-clone...

I did not mean to imply I thought you were wrong, I was completely agreeing with you that it does not appear to be a fruit clone based only on the scores...

But that doesn't mean it isn't, or that it is...

[Andrej Sidorov]

Sidorov
The author of Rybka looked at Stelka's programming and told the world that it had stolen most of his code from an earlier version of Rybka.
Being that was the case as no one doubted his word right after Stelka code was released those who had posted the like as to where it could be found removed the link as it was more or less pirated software.
A few people did manage to download the Streka source before all of this could be determined and apearantly with respect to Rybka author will no longer release that code, so your request to see the source of both programs is now and most likely will be be shown.
I realize that this must be frustrating but Rybkas source was never released to the public and therefore unless the original author releases it, its secrets will remain hidden along with that portion of Strelka that copied Rybka code.
Bill
[/code]

Rogers
Many doubted Rajlich's word after his famous (nods/4) and (plies-3) in Rybka.
Have you any evidence besides 'Vasik says so'?

[ernest]

I don't understand this contradiction, explain me please.

Osipov speaks (you can even find it in Russian )

http://www.talkchess.com/forum/viewtopi ... ial+tables

[Andrej Sidorov]

I don't understand this contradiction, explain me please.

Osipov speaks (you can even find it in Russian )

http://www.talkchess.com/forum/viewtopi ... ial+tables

Osipov speaks:

Vas: ... and even exact data tables are used throughout.

Osipov: If he means the data tables from emater.c file, there is no such data tables in Rybka -- I have checked it. And everyone can check it scanning Rybka's exe. In Strelka there is no other big data tables.

http://64.68.157.89/forum/viewtopic.php?t=18885

[ernest]

...............

See
http://kasparovchess.crestbook.com/view ... 578#p93578

Osipov speaks in the Kasparovchess Forum (11/07/2007), concerning his version Strelka 1.0 Beta:

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

Кроме этого, я обнаружил, что Васик шел тем же путем ? он взял за основу Fruit, переписал его на битборды, и включил таблицы оценок дисбаланса материала от Кауфмана. Некоторые таблицы оценок я похитил из Рыбки.

[Andrej Sidorov]

Some of the tables I kidnapped (ravished) from the Rybka[/b]

So show them, whats he problem?

[ernest]

So show them, whats he problem?

So what more do you want? Osipov himself said that he took the tables from Rybka.
If you want more detail, go ask him (he speaks Russian )

[Michael Sherwin]

So show them, whats he problem?

So what more do you want? Osipov himself said that he took the tables from Rybka.
If you want more detail, go ask him (he speaks Russian )

Strelka 1.0 had the tables. Strelka 1.8 and 2.0 do not have the tables. They were replaced by 'functionals'.

So Vas may be refering to the Strelka 1.0 sources and Osi is saying that they are not in the current sources.

Anyway it seems that both are 'maximizing their injury'.

[Andrej Sidorov]

So lets state that there is no any Rybka tables in Strelka 2.0 B open source.

[Onno Garms]

I think there is one more ingredient which has not been mentioned so far: fine-grained evaluation.

evaluate() uses units of 1/3200 pawn internally. Most weights are not divisible by 32. This is nice to know, but is only useful when you know how to adjust the weights. This secret is not revealed by the Strelka sources. Having only the sources of Strelka's evaluate function is fairly useless unless you want to clone the evaluate function. Appart from material imbalance it contains only well known terms.

I think it's also worthwhile to note what I did not find. AFAIR Rajlich stated that Rybka prunes differently depending on whether the player to move is the same on the current node and the root node or not. Does Strelka do that? I did not find that (but I read search.c less thoroughly then eval.c).