Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A question on Depth reduction?

Author: Pallav Nawani

Date: 12:51:09 08/04/04

Go up one level in this thread


On August 04, 2004 at 15:23:58, Tord Romstad wrote:

>On August 04, 2004 at 14:39:39, Gopikrishna wrote:
>
>>I am doing depth reduction.
>>
>>if (nextdepth > 2)
>>{
>>   if (!Iscapture && !Ispromotion)
>>     {
>>         score= eval();
>>        if (alpha >  score + red_margin[nextdepth])
>>         nextdepth--;
>>     }
>>}
>>
>>So should I have to do full evaluation or just evaluate the material score and
>>pawn structure alone? Because if I do full evaluation for each move it would
>>take plenty of time? Thanks.
>
>I have to disagree with Pallav here.
>
>Reducing the depth based on nothing more than the evaluation is extremely
>risky, and if you don't even do a full eval you can definitely not afford
>it.  Your program will be fast, but it will miss a terrible lot of important
>moves and lines (positionally as well as tactically).  If you really want
>to do something similar to the above code, you need a rather complicated
>static eval which is called at all internal nodes.  You will also need
>many more conditions in addition to !Iscapture and !Ispromotion.
>
>It is not impossible to make something like this work well, but it is
>rather difficult, and should almost certainly be postponed until you
>have a stable and reasonably strong program.
>
>Tord

With a big enough margin it ain't so bad, and besides, it helps a bit when
searching deeper. But yes, it is probably best postponed until you're sure you
have a bug free program.

Pallav



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.