Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: incremental attack tables?

Author: Tord Romstad

Date: 10:13:58 05/05/04

Go up one level in this thread


On May 05, 2004 at 07:48:50, Daniel Shawul wrote:

>On May 05, 2004 at 07:34:55, Tord Romstad wrote:
>
>>Personally, I have always calculated them from scratch at every node (not
>>terribly expensive for me, because the rest of my engine is also dead slow).
>
>  Mine usually searches far lot more nodes than gothmog before i added attack
>tables. But now gothmog searches deeper, i think you use lots of pruning.

I do.  As in all other engines, the search depth displayed by Gothmog doesn't
mean much.  It is the iteration number, nothing more.  Many lines are searched
to depths much lower than the iteration number, and a few lines are searched
much deeper.

In general, I think I reduce and prune more than most others, and extend
less.

>>In my new engine, I am trying to avoid calculating the complete attack tables.
>>Instead, I will try to calculate all attacks to a single square when needed.
>
>   Yes that's waht i did in the previous version. For example in SEE
>i collect attack info for the square and then evaluate. But in evaluate how am i
>going to do it for squares around the king and lots of other things...
>I am definetlely not going back there again.

For king safety you need full attack information only for a few squares
around the king, and not the whole board.  And in some positions (simple
endgames is the most obvious example), there is probably no point in
evaluating king safety anyway.  I think it should be possible to calculate
attack information when needed quickly on the fly, but it is hard to be
sure at the present stage in the development of my engine.

Tord



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.