Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Quiescent v Normal Node Ratio

Author: Peter Kappler

Date: 16:41:45 10/06/99

Go up one level in this thread


On October 06, 1999 at 18:57:17, Peter McKenzie wrote:

>On October 06, 1999 at 18:40:40, Robert Hyatt wrote:
>
>>On October 06, 1999 at 18:16:27, Scott Gasch wrote:
>>
>>>I think it is time we defined some terms here.  Here is what I do:
>>>
>>>depth <= 0  -- qnode.  Note this includes depth=0 horizon nodes.
>>>depth > 0 -- node.
>>>
>>>I am seeing about a 4:1 ration of qnodes to nodes, sometimes a bit less.  I
>>>think that some people are counting ONLY depth=0 nodes as nodes (not qnodes) and
>>>wondering about the ratio of these "nodes" to qnodes (depth < 0).
>>>
>>>Scott
>>
>>
>>Probably correct.  depth=0 nodes are not optional, so they are properly called
>>'leaf' nodes...  any nodes _below_ a leaf is a q-node.  I just count nodes,
>>period, and don't sweat it at all...
>
>I agree, leaf nodes aren't part of the q-search.
>To put it another way: you would still have the same leaf nodes if you didn't
>have a q-search.
>
>The point of counting q-nodes is surely to see how much of the search your
>q-search is taking up.  Counting leaf nodes won't help you here, because it is
>the fullwidth part of the search that 'creates' a leaf node.
>
>I increment nodes when making any move.
>I also increment q-nodes when making a move that is part of the q-search.
>(Although I only do this in debug mode).
>
>Peter


Hi Peter,

It sounds like we count nodes the same way.  What's a typical q-node percentage
for Lambchop?  In Grok its ~15%, which seems to be well below what others are
reporting.

--Peter









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.