Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: hash numbers requested: authors please read

Author: James Swafford

Date: 07:37:29 09/18/02

Go up one level in this thread


On September 17, 2002 at 23:43:17, Will Singleton wrote:

>On September 17, 2002 at 23:11:38, James Swafford wrote:
>
>>On September 17, 2002 at 22:01:42, Will Singleton wrote:
>>
>>>Could you define how you count those values?  People count nodes differently,
>>>and I'm not sure how you count fail highs or lows.  And some folks don't use
>>>fail soft, like me.
>>>
>>>Will
>>
>>Sure.  "Nodes" is the standard "all nodes."  Interior nodes
>>are all nodes not in the quiescence search.
>>
>
>I guess I meant, do you count nodes in makemove, or in calls to search and
>qsearch?  I do makemove, and my contention is that nodes aren't analogous if
>counted the other way.  I think that some folks count nodes in which makemove
>isn't actually called, thus inflating the node count.
>


I don't think I like your definition.  A chess tree is really an
acyclic graph, in which the positions represent nodes and the moves
represent the transitions (arcs).  If your counter is in makemove(),
then you are counting arcs, not nodes.

If you enter Search() with depth=0 and just evaluate the position
and return the score, you've visited a leaf node.  By your definition,
the node count wouldn't increase at all.

Of course, in that case you counted the transition to the leaf
node, but technically that's not a node.  So I think if there's
a significant difference in the two methods, counting in the search
is more accurate (by definition).

That said, I don't see how the difference can be that much.

--
James




>Will



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.