Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Thanks for extracting this!

Author: Vincent Diepeveen

Date: 11:17:22 01/02/06

Go up one level in this thread


On January 02, 2006 at 04:34:05, Vasik Rajlich wrote:

>On January 02, 2006 at 03:11:49, Uri Blass wrote:
>
>>On January 01, 2006 at 16:00:00, Gian-Carlo Pascutto wrote:
>>
>>>On January 01, 2006 at 08:53:35, Uri Blass wrote:
>>>
>>>>I agree but I do not see a reason not to count qsearch nodes and the author did
>>>>not admit that he does not count qsearch nodes and it seems to me based on his
>>>>posts that he counts every node in his search so I was surprised to find out
>>>>that he does not do it.
>>>>
>>>>I did not want to believe GCP that rybka does not count qsearch nodes because I
>>>>assumed that Vasik knows more about rybka and I assume that he does not give
>>>>misinformation but it seems that I was wrong and GCP was right.
>>>
>>>Happy new year!
>>>
>
>Happy new year from me :)
>
>>>Now that you have made this step, it's time to start wondering if Rybka is
>>>really doing something entirely different and new and has such a huge
>>>evaluation, or if we should see Rybka as a Fruit-like engine but with some of
>>>Fruit's weaknesses solved, and if it's really not doing anything really special.
>>>
>>>How much stronger do you think Fruit would get with a better kingsafety and a
>>>better extension system, and some evaluation holes plugged? As strong as Rybka?
>>
>>I believe that fruit can be better than rybka1.2 when it is improved but of
>>course it proves nothing because I also expect rybka to be improved.
>>
>
>Yes, I am sure that Fruit 3.0 and Fruit 4.0 will show that Fabien's approach is
>nowhere near an end.
>
>>It seems to me that rybka also can be improved by a better kingsafety and a
>>better extension system.
>>
>>>
>>>And why is Rybka's endgame so spectacularly bad in some situtions (but fine in
>>>others)? Is Vincents preprocessing theory correct? Was Vasik lazy regarding
>>>those issues?
>>
>
>Ok, not talking about Rybka of course, as I wouldn't like to rob you of the joy
>of figuring this out :), just about pre-processing in general:
>
>1) I am not sure why a preprocessor would play especially badly in the endgame.

This is why i predicted your beta version will be for a long period of time
probably the relative to the competition strongest Rybka version.

Rybka has very little chessknowledge inside in comparision to Diep and Shredder,
so if you start to code a lot, you will encounter the same problems we had.

>2) Amir has explained (you can find this in the archives) that Junior started as
>a preprocessor, but the last traces of its pre-processing were removed around
>'99 or so.

You confuse Amir with Frans Morsch.

Fritz was no longer a preprocessor after 1999, according to Frans himself.

A preprocessor is hereby defined as a program setting up tables to search last
few plies with. That can be 2 ply, that can be 4 ply or it can be entire search.
*where* during the search you initialize those is completely irrelevant.

A professional programmer who used the debugger daily to read assembly code, as
that's what he writes daily for work, claimed not too long ago towards me that
junior still is a preprocessor.

>3) I am not sure why a preprocessor would be more likely to make the sorts of
>weird blunders that Rybka sometimes makes (like 19. .. h3 against Ikarus).

Preprocessors and Fruit clones in bitboards make similar blunders, that is
correct.

Vincent

>Vas
>
>>I do not know the reason but I do not believe vincent's theory.
>>I guess that the reason is that Vasik still did not write the relevant code.
>>
>>Uri



This page took 0.01 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.