Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Some PV thoughts...

Author: Severi Salminen

Date: 00:17:28 11/24/00

Go up one level in this thread


On November 23, 2000 at 17:43:57, Pham Minh Tri wrote:

>On November 23, 2000 at 08:46:36, Severi Salminen wrote:
>
>>Hi!
>>
>>I managed to get my Principal Variation working. Now I have a simple AB search,
>>with iterative deepening and I'm trying to search first the PV moves from last
>>iteration. The strange thing is that I get very little savings in node count
>>when activating the PV sorting. Could the reason be that 1. I evaluate only
>>material and allready have captures tried first. 2.When I place the PV move in
>>front of the queue I place one capturing move further away. I also don't have
>>qsearch yet. How much should "normally" the total node count decrease when I try
>>PV moves first in every ply?
>>
>>Severi
>
>I have seen the same situation in my program. However, I always try the best
>moves from hash table first, and those moves are very similar to pv moves. If
>you have not hash table, something should be wrong. The easiest way to
>experiment, download the TSCP and turn off its pv - the number of nodes will
>increase notabbly.

No, I don't have hash tables yet. But I added some basic square_table evaluation
 and this helped...maybe, as I can't test right now. I also added Qsearch...man,
this is interesting! Do you happen to know how SEE actually works? Should I just
take one square which is attacked more than once and calculate the outcome
statically?

Severi



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.