Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Unsuccessful experiments in move ordering

Author: Robert Hyatt

Date: 11:19:26 12/14/00

Go up one level in this thread


On December 14, 2000 at 09:58:51, José de Jesús García Ruvalcaba wrote:

>On December 14, 2000 at 09:28:02, Robert Hyatt wrote:
>
>>On December 14, 2000 at 04:01:36, Dann Corbit wrote:
>>
>>>I have seen this used successfully:
>>>
>>>Order the moves by the time of the last search.  For ply one (which is
>>>instantaneous) just use generated order.  Then for each subsequent ply, first,
>>>analyse the current pm, then the one that took longest to analyze after that.
>>>Then the next longest.  The one that was fastest should be analyzed last.
>>>
>>>It seems to work very well in practice.  And it is absurdly simple.
>>
>>This only works for the root moves however, which leaves the main part of
>>the tree without help...
>
>Hi Bob,
>have you tried this move ordering in non-root nodes? I do not mean every node of
>the search tree, or course; but perhaps it can be useful in PV-nodes.
>José.


I have never thought about it, because of the cost in terms of memory.  IE
you would have to store a major part of the tree permanently in memory to hold
the various sub-tree counts for each node...  Sounds too expensive to even
think about.



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.