Computer Chess Club Archives




Subject: Re: The Code for the Rybka-Mate-Bug

Author: Vincent Diepeveen

Date: 12:26:18 12/13/05

Go up one level in this thread

On December 13, 2005 at 14:25:00, Chrilly Donninger wrote:

>On December 13, 2005 at 13:47:58, enrico carrisco wrote:
>>On December 13, 2005 at 12:27:31, Chrilly Donninger wrote:
>>>On December 13, 2005 at 11:53:28, enrico carrisco wrote:
>>>>On December 13, 2005 at 09:38:12, Chrilly Donninger wrote:
>>>>>On December 13, 2005 at 09:22:03, Eduard Nemeth wrote:
>>>>>>Hello Chrilly,
>>>>>>thanks for your serching. Nice.
>>>>>>But please search in the code of Fritz 9, I seems Fritz 9 have now 1000 Bugs
>>>>>>(not only 100) ! :-))
>>>>>Who on earth is Fritz? Nobody would be interested if I would post a bug of this
>>>>>noname programme. I was also not asked by the Hydra-sponsor about Fritz. He
>>>>>wanted to know whats special about Rybka. So I have to find out. So far the only
>>>>>special thing I found is the mate-bug.
>>>>>Probably you are right: The secret of Rybka (and Fruit) is the low bugs/second
>>>>>count. Thats a much more interesting figure than the Nodes/Sec.
>>>>I find it odd that Rybka's PV and nps are short and nps seems to be manipulated
>>>>-- either that or sometimes its eval time varies dramatically.
>>>>Maybe it gets its PV from the hash and does not hash in the Q search?
>>>I do not know. Maybe the Rybka team can/want to answer this questions. Generally
>>>I am the wrong person to answer Rybka-Inside questions.
>>It was more speculation than anything.  I highly doubt Vasik will be interested
>>in discussing the particulars of why his nps appears to be manipulated, etc.
>>I am, however, more curious than anything.  I have little doubt that his work is
>>not genuine in its final form.  If I had to speculate further, I would say its
>>origins probably have more in common with Crafty than anything else (e.g., base
>>ideas from -- not clone of course.)  I think a lot of the .EXE size is made up
>>from bitboards precomputed and pattern recognition bitboards for its knowledge.
>>As we have all seen stated earlier, most programs have ideas from others in them
>>and it's clear Rybka has some first class chess knowledge and that coupled with
>>a good search has given it a possible first place among the micros.
>There is the famous quote from Newton: I am dwarf standing on the shoulder on
>giants. Its not possible and even not desirable to invent the whole wheel for
>yourself. Making a somewhat rounder/better wheel is sufficient. In this broad
>sense every programm would be a clone.
>Rybka has Bitboards. But thats mentioned in the Readme File. So I am telling
>here no big news. But I really want to avoid to say some real internals. First
>of all I have not looked on all of Rybkas details. Would be much too much work
>and also boring. I wanted just to have the big picture. And even if I would know
>something important, I could/would not post it. This would be against the rules.

Still it's funny that you claim you score 80% against Rybka, whereas the sheikh
just was interested because he was kicked butt 6.5 - 0.5 by Rybka,
Erdogan operating Rybka.

Several games of that the sheikh played with a 32 node monster, though it
was unclear whether it had 32 or 64 FPGA processors :)

In any case, the bugs per second of rybka is always higher than that of Hydra in
software, because i assume Rybka gets more nodes a second in software than Hydra
gets in software :)


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.