The Secret of Rybka Success

[fern]

What do you think it is the key (mainly) of Rybka incredible results?

a) Vas has implemented several new decisive tricks in the evaluation realm that has given his program the extra boost
b) Vas has implemented new, betters search routines.
c) Vas has simply improved in both.
d) the bastard signed an agreement with Satan.

My guess:
evaluation superiority

Best to all
fernando

[AdminX]

What do you think it is the key (mainly) of Rybka incredible results?

a) Vas has implemented several new decisive tricks in the evaluation realm that has given his program the extra boost
b) Vas has implemented new, betters search routines.
c) Vas has simply improved in both.
d) the bastard signed an agreement with Satan.

My guess:
evaluation superiority

Best to all
fernando

All of the above, plus great understanding of imbalances.

[Dirt]

What do you think it is the key (mainly) of Rybka incredible results?

a) Vas has implemented several new decisive tricks in the evaluation realm that has given his program the extra boost
b) Vas has implemented new, betters search routines.
c) Vas has simply improved in both.
d) the bastard signed an agreement with Satan.

My guess:
evaluation superiority

Best to all
fernando

From what Larry K. posted in the Rybka forum, is sounds like Vas found a way to automatically generate evaluation parameters from a game database. Larry was posting that he was tweaking some of the parameters that seemed wrong.

[Kapaun]

I believe that he has implemented all the things that proved to be working - even if chess knowledge advised otherwise.

[royb]

I may be wrong, but I thought Hans Berliner was doing something like this in the 80s with HiTech. Hopefully someone here can confirm what Hans was doing and also if it is in some way similar to what Vas and Larry have done with Rybka.

Roy

[Allard Siemelink]

I used to do that for Bright and although I got some interesting results,
the strength of the engine did not seem to improve when I 'enhanced' the evaluation with what I found.
Perhaps I should give it another try?

From what Larry K. posted in the Rybka forum, is sounds like Vas found a way to automatically generate evaluation parameters from a game database. Larry was posting that he was tweaking some of the parameters that seemed wrong.

[Alessandro Scotti]

d) the bastard signed an agreement with Satan.

Doesn't matter. Can offer pawn odds and still win his soul back!

[Spock]

Looking at the article here:
http://www.rybkachess.com/docs/Rybka_Englisch.PDF

I wonder if his previous work on artificial intelligence is a factor....

[RVisitor]

We can only guess at the heuristic algorithm Vasik uses for Rybka. Below is a link to a paper where I offer my opinion.
http://mysite.verizon.net/vzesz4a6/current/id23.html

[fern]

Your opinion about this are so sensate I guess some of them, if no all, has been more or less used since some time.
I remember I wrote a paper about this and sent it to a respected programmer. My ideas were not as you, but has a common factor; the idea of looking the chess game as it is, as process, a historic thing going on and getting sense along it.
In my paper a knight was not strong because he was posted in e5, but because it was adding strength along an previously existent pattern of pressure. By example, he was adding another piece-action over an enemy piece.
And so on.
I tell you this because I have realized that the abyss between sensate ideas like these or others AND the execution of them in a code is immense. There lyes the problem, in the know-how issue. Probably there lyes Vas excellence.

My best
Fernando