Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: significant math

Author: Steffan Westcott

Date: 15:58:54 11/20/02

Go up one level in this thread


On November 20, 2002 at 17:58:33, Sune Fischer wrote:

>>Sune,
>>
>>Just so I can address your question properly, what precisely do you mean by the
>>term 'book keeping' ?
>
>
>Well it's not easy to explain, actually, but let me give it a try.
>
>It's about not generating the same move twice, it's about not wasting a lot of
>energy avoiding to generate the same move twice.
>
>I can understand the principle if you do MVV/LVA, the patterns are straight
>forward. So you go through the list until you find a move and then try that.
>
>I admit I can't see a fast way of doing this, it appears you need to ask ~25
>questions to run through this list. Suppose there are no bishops on the board
>and that black has no queens. So you have some pointers you point to something
>so you avoid all Bx?,?xB and ?xQ questions?
>How do you do this?
>
>Now the non-captures, have you saved the calculated attack boards, or are those
>recalculated on the fly, again?
>

As I've said somewhere else on this monster thread, I generate some auxiliary
state when a search node is first visited. This state is _not_ portable between
search nodes. Recalculation for all but the most trivial of expressions is
minimised.

Generating the movelist first is an effective way of reducing the number of
moves to consider. Going from the set of around 4000 all possible moves to
around 35 pseudo legal moves is not a bad first step! As each move is searched,
the corresponding bit is reset in the 'movelist' data - Again, see another post
of mine for detail on this. If the standard move heuristics fail to suggest
anything useful, only then does SME take over to attempt a sensible move
suggestion (without the benefit of any searching). Note it only is required to
find one move at a time. It is usually not required to use SME throughout the
whole movelist - SME is used to suggest a 'few' good move candidates (usually
good captures as found by traditional SEE methods). The type of search node
affects use of SME too.

Cheers,
Steffan



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.