Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Computer chess schools of thought

Author: Richard Pijl

Date: 06:28:49 07/01/02

Go up one level in this thread


On July 01, 2002 at 08:57:20, Sune Fischer wrote:

>On July 01, 2002 at 07:50:12, Richard Pijl wrote:
>
>>On July 01, 2002 at 07:18:45, Sune Fischer wrote:
>>
>>>On July 01, 2002 at 07:07:10, Richard Pijl wrote:
>>>
>>>>On July 01, 2002 at 06:50:51, Sune Fischer wrote:
>>>>
>>>>>On July 01, 2002 at 06:34:31, Richard Pijl wrote:
>>>>>
>>>>>>>I presume the reason for this is the slow InCheck implementation, so the idea is
>>>>>>>to call the heavy functions as little as possible.
>>>>>>
>>>>>>And search less nodes as you do not have to search the check evasions. I think
>>>>>>that is the real winner.
>>>>>
>>>>>Oh well I don't do that in qsearch. It just means that I might evaluate a
>>>>>position where the king is in check, but that's still better than to evaluate a
>>>>>position where the king has been captured, IMHO. :)
>>>>>
>>>>
>>>>In that case you could just assign a very high material value to a king, so you
>>>>can determine a captured king by the material balance and skip evaluation with a
>>>>fail low ... Then you have no need for calling InCheck in qsearch at all
>>>
>>>But my InCheck is fast (at the expense of slow make-/unmakemove), so it's no
>>>problem for me.
>>>And I don't see how it solves Craftys problem, you still might evaluate
>>>positions where the king is in check.
>>
>>Yes, but if you're not doing check evasions in qsearch, it doesn't matter, does
>>it?
>
>I misunderstood, I thought the idea of check evasions in qsearch was _not_ to
>evaluate while in check, since technically it cannot be considered a quite
>position.
>
That is correct. What I meant was that when you're not calling InCheck from the
qsearch (and not doing checkevasions) you could save a bunch of nodes. Of course
the resulting position is not quiet, but on the otherhand: If you are in check
in qsearch and you cannot evade the check by a capture move, chances are that
you cannot improve the score anyway.

I'll do some tests with disabling check evasions this evening. As I recently
implemented SEE and prune in qsearch based on SEE, things may not be as bad as
they were before.

Richard.



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.