Computer Chess Club Archives




Subject: Re: Intelligent software, please

Author: Christophe Theron

Date: 17:37:39 11/26/01

Go up one level in this thread

On November 26, 2001 at 17:10:37, Steve Maughan wrote:

>>>I never tried the idea to evaluate based on depth so from experience I
>>>can not say the idea is good or bad. The downside of your above idea I
>>>can imagine is that for instance a Nc7xa8 rook capture is rejected at a
>>>15/16 ply depth while it is really needed to take the rook.
>Hmmm, not sure that this is a problem.  It will only be rejected at 15/16 ply if
>the knight really cannot get out.
>>I have considered this, but I'm not going to do it (not in the near future)
>>because of problems with the hash table. The hash table can store scores and
>>bounds found by different iterations, and if the positional weights are not
>>consistent from iteration to iteration then the scores stored in the hash
>>tablesare not going to be consistent either.
>>I think it's a major problem. It would maybe not hurt much in the middlegame,
>>but I fear it would kick badly in the endgame.
>Good point - maybe a better approach would be to base it on the speed of play
>i.e. classify some terms as speculative and tune them down for fast play - up
>for slow play.  This way you'd have the same weights from iteration to
>iteration.  Just a thought!

You are right, but suddenly the idea sounds more and more tricky to implement.
It was simple when it was about search depth (disable some terms when the
position to evaluate is deep enough), it's not anymore if you have to calibrate
on the speed of the computer and the time controls...


This page took 0.06 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.