Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Thanks to the following for their assistance on move ordering...

Author: Dann Corbit

Date: 12:29:41 05/04/01

Go up one level in this thread


On May 04, 2001 at 15:25:39, Will Singleton wrote:

>On May 04, 2001 at 14:27:54, Dann Corbit wrote:
>
>>Christian Söderström
>>Gian-Carlo Pascutto
>>Ulrich Tuerke
>>Robert Hyatt
>>
>>And even David Rasmussen, because his number was of value, even if his
>>explanation was a bit lacking.
>>;-)
>>
>>It seems that move ordering is something of a black art.
>>
>>Everyone does it a little differently.
>>Everyone even *counts* it a little differently (so we have to think clearly to
>>see if we are comparing apples to oranges).
>>
>>And yet, it is one of the most important facets of the search.  The performance
>>jump in going from 70% correct to 90% correct will be enormous.
>>
>>I also went through the information in Ernst Heinz's excellent book, and
>>Pepito's code.
>
>
>I was interested in the subject, and was going to run it.  But I have to admit
>that I couldn't quite get what you were asking, so I blew it off.
>
>I've worked quite a lot on move ordering, and would be interested in any
>conclusions you might have.

The conclusions are fairly simple:
1. It is incredibly important to get it 'close' to right.  If you don't average
90%, you are taking a punch in the chops.
2. Different methods seem to all work just as well.
3. Try them all, including hybrids, and use whatever works best.



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.