What do you mean with being thorough? I fear this is without experience from a frog's perspective. Look, you give a lot of numbers and perhaps this gives to yourself already partially the impression that you are doing thorough science but this is a delusion.Zach Wegner wrote:Hey Mike, sorry for the late response. I wanted to be thorough on this, plus school is getting pretty hectic right now. Anyways...Michael Sherwin wrote: Sorry Zach. Guess that I am just too exhausted to read with
comprehension. Taking care of a parent with Alzheimer's will do that
to a person.
1) So there is a lot of Fruit in both engines that is the same between them.
2) There are some small amount of things that Rybka and Strelka do
differently that are not in fruit.
3) And there are only a few things that Strelka takes from Fruit that
are not to be found in Rybka.
4) There are things in Strelka besides the material tables that are
also in Rybka, but are not in Fruit.
Are all these statements true? If so then please elaborate on point 4
as this would be the proof that Osipov took more from Rybka than just
the material tables. Are the move generators nearly identical? I read
that they are not the same.
Anyway, I am going on vacation from this madness for a while. Though I
am still interested in an answer.
Every one of 1-4 is true to some extent. I'd estimate the rough percentages like this:
1) Rybka/Strelka, in terms of algorithms (not implementation), is
maybe 80-90% Fruit. Of course the code is fairly different in a lot of
cases, because Rybka is bitboard, but the chess-playing logic is very
similar. I guess I'd say that the Rybka 1 eval is 90% Fruit, and the
Rybka 1 search is 80% Fruit.
2) There are very few differences between Rybka and Strelka, maybe a
few percent. These are very small details too. Maybe 1-2% of Strelka is different than Rybka, and about 3/4 of that is not from Fruit.
3) Of the 1-2% mentioned in 2, this is maybe a quarter.
4) This is rather large, about the opposite of 1). In virtually all of the places where Rybka and Fruit differ, Strelka is the same as Rybka. While Rybka 1's eval is semantically very close to Fruit's, the implementation is translated to bitboards, and here, Strelka is the same. Same with the search, Strelka's futility pruning is like Rybka's, etc.
There is more to the analysis, of course. I'm trying to finish up my webpages (slowly), so I can have some more detailed info. Feel free to ask any other questions about this, though...
It is surprising that I can easily read and understand what you are giving here and I never was the youngest nor oldest nor any programmer ever at all.
What springs to mind is that you throw around with numbers and percetages but you give no scales for what these numbers should stand. They are meaningless at least in that shortness.
I dont want to teach you anything because I dont want to support this hate motivated labor against Vas but of course if you had results, fair and thorough, then I would easily accept them. I'm not a witch doctor who selects the data that he prjudiced and what he accepted. If data is sober I accept it but this above is not sober at all.
The problem is a) you have no scale b) no comparisons with the rest of the field, you want to imply for lay that normally the engines would differ so that you perhaps would get 1% or 1,5% similarity. All this is of course unknown to me because I am not a computer scientist. But I doubt that this is what in real would come out. Just for the debate I have already pretended as if the progs all had stuff similarities of 95%, stuff that was taken from others. If this is true then where is your argument?? Or are you just another bean-counter who presents data without any background and meaning?
Nothing against you. You can be talented, but you cant have studied all these topics yet? Questions of methodology. Science of science? What then would allow you to judge what has for how many percentagr probably what impact.
Look, if you have all that ready after also advices from experts THEN your labor will only begin. Because then you must decide how much of what you formerly thought as relevant has now any impact at all. Be cautious. Perhaps others want to use you as proxy. Because you are young, your risks are not as high as theirs.
Only if after all this you can calculate if it's really worth to go public and to smear the good name of someone else.
But this here means nothing at all.
Let me still ask a question. I wished to see if you really are an American guy who typically never fears anything and is proud to speak about what he thinks is the truth.
I missed that in detail. I understood that you presented all such data on rybkaforum and Vas didnt answer but someone, or was it Vas told you that on that base what you had presented that you must be able to create a real killer machine. Who was it who told you this?
Well this is what I can comment on the message above. Being no programmer it might well be that the real difficulties are somewhere else than there where I saw a problem, but what does that matter. I could be your grandfather and I wish you really all the best and that computerchess might be interesting for still some decades. If I got a wish free from you then it's that you might become a bit more careful in your public statements.
Look, D. Corbit must have a different style. He reported that he even exchanged emails with Vas and he told him a bit more about the second clone affair and Dann say he believes Vas. He certainly didnt see the final proof either. But this is a difference to what you write, no?
I can only tell you what I already told Anthony, when he appeared a bit often here, try to get your studies done, that is the most important. Because your perspective is changing. That here is bringing you off your own paths. Seeking proof against someone is always negative labor. It might be that you succeed but still weont have true success, simply because people wont like it. They would fear you. Also think about the rlevance of CC in computer sciences.
Well, blabla, just tell me some comparisons with other engines.
Rolf