what is the secret of rybka

Discussion of chess software programming and technical issues.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Uri Blass
Posts: 8604
Joined: Wed Mar 08, 2006 11:37 pm
Location: Tel-Aviv Israel

Re: what is the secret of rybka

Post by Uri Blass » Tue Aug 07, 2007 1:31 pm

<snipped>
Daniel Mehrmann wrote: But this says not that her search is weak compared to other top engines. Its more or less equal compared with top engines. Vas isn't using unknown or new search technics. He just tuned the well known public technics.

Best,
Daniel
How do you know that vas does not use search technics that other top engines do not use?

I do not know and I have no reason to assume that the only reason that rybka is better than Junior or Fritz is better evaluation.

Uri

RVisitor

Re: what is the secret of rybka

Post by RVisitor » Wed Aug 08, 2007 1:52 pm

Perhaps the ideas contained in the paper are best presented to the International Computer Games Association (ICGA) to determine whether or not they have any potential.

kaustubh

Re: what is the secret of rybka

Post by kaustubh » Wed Aug 08, 2007 2:08 pm

hello daniel,
I don't think i am writing non-sense. I ask what is the secret of rybka because I think vasik is using some new search or evaluation techniques ,there are experienced programmers here so they must know new techniques which i want to understand.
With regrads,
kaustubh

nczempin

Re: what is the secret of rybka

Post by nczempin » Wed Aug 08, 2007 2:20 pm

kaustubh wrote:hello daniel,
I don't think i am writing non-sense. I ask what is the secret of rybka because I think vasik is using some new search or evaluation techniques ,there are experienced programmers here so they must know new techniques which i want to understand.
With regrads,
kaustubh
If we "experienced programmers" knew his "secret", it wouldn't be a secret any more. We would simply adapt our own engines (or at least those that are writing competitive engines, not like me using substandard technology and techniques :-)) and the gap would disappear.

So, since the gap is still there, what would be a logical conclusion?

.. yes?

The only reason we are mocking you is because it seems to be so _very_ obvious that we DON'T know the "secret". So there's really no point in asking.

In addition, if I knew the secret, I probably wouldn't post it here, unless I had some kind of axe to grind with Vas.


Additionally: If all you want is to speculate among with the rest of us what the secret really is: The point is that all the speculations have already taken place, as well as all the interesting discussions. It seems like you arrived at the wrong time to ask such questions.

Michael Sherwin
Posts: 3046
Joined: Fri May 26, 2006 1:00 am
Location: WY, USA
Full name: Michael Sherwin

Re: what is the secret of rybka

Post by Michael Sherwin » Wed Aug 08, 2007 2:51 pm

The funny thing is that there is no secret. Vas, when he was posting here said exactly what he was doing. I even posted an example that highlited that Rybka was doing what Vas said that he was doing. It is really quite simple. Vas divided a rather standard eval score by 2 and then added score adjustments similar to Kaufman's values. The large table of data that Rybka uses for this purpose is data that Vas gathered from a database of games, just like Kaufman did it.

I posted this quite awhile ago, long before Strelka came out. The author of Strelka has done a very good job at verifying what I posted, even though he most likly never read my post!
I hate if statements. Pawns demand if statements. Therefore I hate pawns.

kaustubh

Re: what is the secret of rybka

Post by kaustubh » Wed Aug 08, 2007 3:05 pm

Thank you for your reply , can you give any links. :D :D :D

Uri Blass
Posts: 8604
Joined: Wed Mar 08, 2006 11:37 pm
Location: Tel-Aviv Israel

Re: what is the secret of rybka

Post by Uri Blass » Wed Aug 08, 2007 3:20 pm

Michael Sherwin wrote:The funny thing is that there is no secret. Vas, when he was posting here said exactly what he was doing. I even posted an example that highlited that Rybka was doing what Vas said that he was doing. It is really quite simple. Vas divided a rather standard eval score by 2 and then added score adjustments similar to Kaufman's values. The large table of data that Rybka uses for this purpose is data that Vas gathered from a database of games, just like Kaufman did it.

I posted this quite awhile ago, long before Strelka came out. The author of Strelka has done a very good job at verifying what I posted, even though he most likly never read my post!
You talk about rybka beta but rybka improved from that time.
I can add that I dislike these questions.

If people really know the secret of rybka they could write better program.

Claiming that rybka based on some addition to fruit-toga does not help
even if it is correct because programmers usually do not know what is the secret of toga inspite of the fact that it is an open source program(otherwise they could easily write something better).

Edit:I can add that I think that only one person know the secret of rybka so I suggest the poster to ask vasik directly and not to ask us.

If the poster pay Vas enough money maybe Vas can tell him the secret.

Uri

User avatar
Daniel Mehrmann
Posts: 855
Joined: Wed Mar 08, 2006 8:24 pm
Location: Germany
Full name: Daniel Mehrmann

Re: what is the secret of rybka

Post by Daniel Mehrmann » Wed Aug 08, 2007 3:25 pm

Uri Blass wrote:
Michael Sherwin wrote:The funny thing is that there is no secret. Vas, when he was posting here said exactly what he was doing. I even posted an example that highlited that Rybka was doing what Vas said that he was doing. It is really quite simple. Vas divided a rather standard eval score by 2 and then added score adjustments similar to Kaufman's values. The large table of data that Rybka uses for this purpose is data that Vas gathered from a database of games, just like Kaufman did it.

I posted this quite awhile ago, long before Strelka came out. The author of Strelka has done a very good job at verifying what I posted, even though he most likly never read my post!
You talk about rybka beta but rybka improved from that time.
I can add that I dislike these questions.

If people really know the secret of rybka they could write better program.

Claiming that rybka based on some addition to fruit-toga does not help
even if it is correct because programmers usually do not know what is the secret of toga inspite of the fact that it is an open source program(otherwise they could easily write something better)

Uri
I don't think Michael was refering to Toga/Fruit. He just talks about a standard evaluation and what Vas could be done.

Best,
Daniel

Michael Sherwin
Posts: 3046
Joined: Fri May 26, 2006 1:00 am
Location: WY, USA
Full name: Michael Sherwin

Re: what is the secret of rybka

Post by Michael Sherwin » Wed Aug 08, 2007 3:40 pm

Uri Blass wrote:
Michael Sherwin wrote:The funny thing is that there is no secret. Vas, when he was posting here said exactly what he was doing. I even posted an example that highlited that Rybka was doing what Vas said that he was doing. It is really quite simple. Vas divided a rather standard eval score by 2 and then added score adjustments similar to Kaufman's values. The large table of data that Rybka uses for this purpose is data that Vas gathered from a database of games, just like Kaufman did it.

I posted this quite awhile ago, long before Strelka came out. The author of Strelka has done a very good job at verifying what I posted, even though he most likly never read my post!
You talk about rybka beta but rybka improved from that time.
I can add that I dislike these questions.

If people really know the secret of rybka they could write better program.

Claiming that rybka based on some addition to fruit-toga does not help
even if it is correct because programmers usually do not know what is the secret of toga inspite of the fact that it is an open source program(otherwise they could easily write something better).

Edit:I can add that I think that only one person know the secret of rybka so I suggest the poster to ask vasik directly and not to ask us.

If the poster pay Vas enough money maybe Vas can tell him the secret.

Uri
You are correct, Vasik stopped talking after the beta came out, so, we have no idea what his great mind might have thought of since.

And you are partially right about Toga being an enigma, despite it being open source, but, not to everyone. There are a few that have learned from the Toga sources and that would explain the advancement of quite a few top level engines. The problem is, is that they can not take the chance of admitting it, because, they would risk running afoul of the GPL, not to mention, public opinion.
I hate if statements. Pawns demand if statements. Therefore I hate pawns.

kaustubh

Re: what is the secret of rybka

Post by kaustubh » Thu Aug 09, 2007 6:26 am

nczempin wrote:
kaustubh wrote:hello daniel,
I don't think i am writing non-sense. I ask what is the secret of rybka because I think vasik is using some new search or evaluation techniques ,there are experienced programmers here so they must know new techniques which i want to understand.
With regrads,
kaustubh
If we "experienced programmers" knew his "secret", it wouldn't be a secret any more. We would simply adapt our own engines (or at least those that are writing competitive engines, not like me using substandard technology and techniques :-)) and the gap would disappear.

So, since the gap is still there, what would be a logical conclusion?

.. yes?

The only reason we are mocking you is because it seems to be so _very_ obvious that we DON'T know the "secret". So there's really no point in asking.

In addition, if I knew the secret, I probably wouldn't post it here, unless I had some kind of axe to grind with Vas.


Additionally: If all you want is to speculate among with the rest of us what the secret really is: The point is that all the speculations have already taken place, as well as all the interesting discussions. It seems like you arrived at the wrong time to ask such questions.
Knowing some algolrithm is a different from implementing that algolrithm in your engine. Vasik said he spend 2500 hours on rybka. So there must be some new techniques which are hard to implement but are good, like mtd(f)

Post Reply