Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Move ordering scheme

Author: Andrew Wagner

Date: 08:44:25 03/16/04

Go up one level in this thread


On March 16, 2004 at 09:33:17, Uri Blass wrote:

>On March 16, 2004 at 09:20:48, Andrew Wagner wrote:
>
>>I currently use the following scheme to order my moves:
>>1.) Hash move
>>2.) Killer moves
>>3.) Capturing the piece that just moved
>>4.) Other captures
>>5.) History heuristic
>>6.) Moves that land closest to the center
>>7.) All other moves
>>
>>(I think I got that right, that's from memory).
>>
>>I currently get 85-95% first-move fail highs right now, which I'm pretty happy
>>with. However, I've noticed two things:
>>
>>     1.) Often, if the move ISN'T a fail high, it doesn't fail until 10 or 15
>>moves into the move list.
>>
>>     2.) I'm searching a LOT of nodes! Most of the time I'm searching more nodes
>>than crafty by a factor of at least 10.
>>
>>Any suggestions?
>
>1)you should use good captures before killer movers when killer moves do not
>include captures.
>

Ok, I'll take a look of that.

>
>2)What do you mean when you say that you search more nodes than Crafty?
>Do you search more nodes to get the same depth or search more nodes at the same
>time?
>
>In the first case there is a problem to compare because you may do differemt
>extensions and reduction.
>
>In the second case it seems that your engine is very fast(I do not know about a
>single engine that is 10 times faster than Crafty in nodes per second).
>
>Uri

If Crafty searches 5000 nodes to get to a certain depth in a position, my engine
searches at least 50000 nodes. Which means of course it is slower to reach that
depth, even if the nps were the same. Thanks!



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.