a response to anthony Cozzie opinion on rybka

[bob]

I really had appreciated if an expert like you would have given a better defense for Bob against some "Horst" who simply addressed impolitely. More, he obviously even confused Strelka with Rybka...

I did, but it got deleted by the moderators every time.

Tony

Tony - referring to some members as firstly stupid people asking stupid questions and then as silly people asking silly questions was not acceptable. That's why those posts were removed.

Regards, Graham.

At times that is a reasonable follow-up. "stupid question" or "silly question" is perfectly acceptable. And telling someone they are silly for asking a question is not outside the charter either. That's hardly an attack on someone's character. If we can't say things like "you are silly for asking such a thing" then there is very little that can be said here...

[bob]

Perhaps you would be kind enough to explain the circumstances under which you might expect an open source program to be stronger than a closed source commercial program.

you imply, that every commercial programmer uses the public code just as he wants. there were and still are some commercial programs, that are weaker than fruit for instance.

if you want to strictly seperate rybka 1.0 and strelka, then one or maybe even Mr. Hyatt could examine the strelka source, but as he already said, he won't do that probably. my conclusion is, that crafty has been just a toy to implement some new stuff for quite a while now and not really an object of improvement.

how do you want to make progess in playing strength, if you don't even test the prog against others and so on. that's of course up to Hyatt, but constantly rallying against the commercial programers, who don't share their new ideas with the comunity is strange.
as Cozzie stated, there isn't even anything "new" in rybka/strelka - so Hyatts has no point at all.

I'd bet I have played more games against other programs than everybody else posting here combined. I play about 30,000 games a day, _every_ day. Sometimes more. So I guess I don't even begin to comprehend what you mean by "if you don't even test the program against others..." Do you have any idea what you are saying? Didn't think so. I've explained my testing approach here several times. I use several other programs, including Fruit, Glaurung, Arasan, gnuchess, and others. And I play more than "just a few" games. So please pay more attention to what is going on before you make statements that are completely false and/or misleading.

BTW, I doubt you would be the person I would ask if I wanted to know if there was anything new in Strelka. In fact, other than the people on the Crafty team, there is nobody I would ask, Certainly not a commercial programmer who would have a vested interest in not revealing any novel ideas that might be in Strelka but not apparent to casual observers. If I become interested, I will look for myself...

[Nid Hogge]

http://rybkaforum.net/cgi-bin/rybkaforu ... 7#pid42407

Vasik Rajlich :

"Sorry, this is just nonsense.

In fact, I've never even profiled Rybka and spend zero time on optimization. This should be quite obvious from the sources - there are no unrolled loops or other arcane constructions, assembly sequences which don't map to C, etc.

It just doesn't seem like a productive area, especially long-term.

A few comments for any bitboard fanatics who might be browsing here:

1) I typically put elegance and simplicity before speed. Don't look for too much meaning at the low level. Someone like Gerd Isenberg could probably speed Rybka up by 10-15% without crossing over into any really hard-core optimization.

2) I've always just used plain Crafty-style rotated bitboards, and haven't yet managed to try anything else. My intuition is that the magic number approach (which wasn't around when I started) would be a little bit better. This may be doubly true for Rybka, as I suspect that she pollutes the cache more than most engines. If I started today, this is probably what I'd go with."

Vas

it would probably take far more than an afternoon to wade through any program and verify that there is "nothing new at all"

Indeed.

Time to take another look, anthony

[frosch]

I am sorry for my wrong statement about not testing against other programs. it was based on a posting of you, where you reportet, that a new crafty version for the WC has not been tested against other programs.
I probably got something wrong or it was an exception.

[Michael Sherwin]

I am sorry for my wrong statement about not testing against other programs. it was based on a posting of you, where you reportet, that a new crafty version for the WC has not been tested against other programs.
I probably got something wrong or it was an exception.

Bob indicated (iirc) that the version for the WC is still in the design phase. With a lot of modification being done. The testing/tuning phase against many engines will come later.

[Albert Silver]

http://rybkaforum.net/cgi-bin/rybkaforu ... 7#pid42407

Vasik Rajlich :

"Sorry, this is just nonsense.

In fact, I've never even profiled Rybka and spend zero time on optimization. This should be quite obvious from the sources - there are no unrolled loops or other arcane constructions, assembly sequences which don't map to C, etc.

It just doesn't seem like a productive area, especially long-term.

A few comments for any bitboard fanatics who might be browsing here:

1) I typically put elegance and simplicity before speed. Don't look for too much meaning at the low level. Someone like Gerd Isenberg could probably speed Rybka up by 10-15% without crossing over into any really hard-core optimization.

2) I've always just used plain Crafty-style rotated bitboards, and haven't yet managed to try anything else. My intuition is that the magic number approach (which wasn't around when I started) would be a little bit better. This may be doubly true for Rybka, as I suspect that she pollutes the cache more than most engines. If I started today, this is probably what I'd go with."

Vas

it would probably take far more than an afternoon to wade through any program and verify that there is "nothing new at all"

Indeed.

Time to take another look, anthony

Oh boy....

"When Larry joined me in 2007, we went back and worked on the eval again, under a completely different philosophy. The Rybka 2.3.X versions were a sort of early prototype of this method. I was quite happy with these steps and now they've been taken much further, to the point that Rybka is nearly unrecognizable."

Albert

[Nid Hogge]

?

This was exactly what I was saying already

http://www.talkchess.com/forum/viewtopi ... f07#172579

you have to remember that Strelka is based on Rybka 1.0 32-bit beta. Many things could have benn changed since, including eval, search and whatever. so the statements are only in line and applied for strelka. that's how I see it.

Just the way I thought it is.

Seeing all those "so Rybka is has nothing special after all" and "it's doing X, it's bad in eval, strong in search" type of posts made me jiggle. But now that's a good thing, so that there's nothing really spectacular that can be copied by other engines. At least until Senior Osipov decides to strike again.. Hmm.

[bob]

I am sorry for my wrong statement about not testing against other programs. it was based on a posting of you, where you reportet, that a new crafty version for the WC has not been tested against other programs.
I probably got something wrong or it was an exception.

Several years ago the only testing we were doing was on the internet chess club. But since Crafty was started in 1994, it has always played tens of thousands of games every year, which is one reason you have never seen it crash/hang in a tournament. For the last couple of years, we have been playing an incredible number of highly-controlled matches on our clusters, which makes it possible to play 20,000 high-quality games in under a day... And we are running these tests evry single day.

[Tony Thomas]

I am sorry for my wrong statement about not testing against other programs. it was based on a posting of you, where you reportet, that a new crafty version for the WC has not been tested against other programs.
I probably got something wrong or it was an exception.

Did you ever play the games against Crafty as you said you would?

[Mark]

I am sorry for my wrong statement about not testing against other programs. it was based on a posting of you, where you reportet, that a new crafty version for the WC has not been tested against other programs.
I probably got something wrong or it was an exception.

Several years ago the only testing we were doing was on the internet chess club. But since Crafty was started in 1994, it has always played tens of thousands of games every year, which is one reason you have never seen it crash/hang in a tournament. For the last couple of years, we have been playing an incredible number of highly-controlled matches on our clusters, which makes it possible to play 20,000 high-quality games in under a day... And we are running these tests evry single day.

That's a lot of games! How do you use all of this data to improve Crafty? Do you try to categorize the losses somehow and look for areas where the eval needs help or what? It seems like there would be too much data to do anything manually. Is there any automated review of the games?