Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: How to extend if only one reply?

Author: Robert Hyatt

Date: 10:11:21 01/24/01

Go up one level in this thread


On January 24, 2001 at 13:03:21, Miguel A. Ballicora wrote:

>On January 24, 2001 at 03:11:56, Severi Salminen wrote:
>
>>>I didn't do this in Cray Blitz so there is an alternative.  Do the normal
>>>search, but each time you search a legal move, inc a counter.  When you are
>>>done searching all moves, if the counter == 1, there was only one legal one
>>>to search.  Search it _again_ but one ply deeper...
>>
>>Sounds time consuming...I hate all these researchs :) I _do_ use a legal move
>
>It should not be, I think. It is like an internal iterative deepening.
>Moreover, it could be faster! couldn't be?
>
>>counter (for other purposes) but I believe it is faster to generate only legal
>>"check evasive" moves as I allready pass the "being in check" information to the
>>next ply.
>
>One question, you people do extend when there is a single move. This is an
>extra extension besides "check extensions"?
>In other words, if king is in check --> depth+=1 and if there is only one
>reply --> depth+=2 (assuming no fractional increments) ?
>
>Regards,
>Miguel
>
>
>>
>>Severi


I get away with doing two extensions because of a minor oddity in my search.

I decided that due to my very simple q-search, I didn't want to make a move
at the last ply that would check my opponent, and hand that position over to
my q-search.  What I do, therefore, is that I extend when I _check_ the other
side, rather than waiting until I get to the next ply and extending when I
notice I am in check.  At the next ply I notice I am in check, call the legal-
move generator, and if it produces only one move, I extend there for 1 reply.

Note that my 'check' extension is 1.0 plies, but the 1-reply extension is
currently .75 plies.  When I start the search, I start depth off at 1.5, which
means the first 2 1-reply (or any other fractional extension) extensions will
actually extend a full ply, but the third won't.  From then on, out of a group
of four 1-reply positions, three will extend and one won't...



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.