Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A note about Rybka's "speed" and variation display

Author: Daniel Shawul

Date: 20:18:06 12/05/05

Go up one level in this thread


On December 05, 2005 at 16:46:57, Joachim Rang wrote:

>On December 05, 2005 at 10:37:29, Vasik Rajlich wrote:
>
>>On December 05, 2005 at 07:19:25, Gian-Carlo Pascutto wrote:
>>
>>>It seems to me that Rybka's seemingly slow speed is caused by not counting
>>>quiescent nodes as search nodes.
>>>
>>>You can observe that Rybka's nodes per second drops dramatically in positions
>>>where there are lot of forced check sequences.
>>>
>>>I am actually pretty sure of this because of 2 reasons:
>>>
>>>1) it never displays the parts of the variation coming out of the quiescent
>>>search
>>>2) it can solve mates with an unrealistically small nodecount
>>>
>>>--
>>>GCP
>>
>>Very clever :) I also like to crack engine q-searches. You will find Rybka's a
>>bit more difficult than most. In fact, I dare say impossible. ;)
>>
>>Re. nps, I am thinking about artificially doubling it. Kind of like what Junior
>>did with their depths. I got already several emails asking what is wrong with
>>Rybka or if there was an installation problem ..
>>
>>Vas
>
>
>It is also possible that Rybka does all kind or quiescence task in eval. Since
>it is a bitboard engine you can relatively easy do all kind of "captures" in
>eval.
>
   I doubt bitboards will make life any easier. Ok detecting pins may be
simple,but doing a full board swap off is as hard as a piece list
representation.
daniel

>regards
>
>Joachim



This page took 0 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.