Author: Uri Blass
Date: 04:47:53 09/13/04
Go up one level in this thread
On September 13, 2004 at 06:24:05, Gerd Isenberg wrote:
>On September 13, 2004 at 05:46:12, Uri Blass wrote:
>
>>On September 13, 2004 at 05:27:57, Gerd Isenberg wrote:
>>
>>>On September 13, 2004 at 00:46:30, Michael Henderson wrote:
>>>
>>>>Hello, can anybody confirm these definitions I have for nodes? I think they are
>>>>right but I like to make sure :)
>>>>
>>>>(depth = 2) //pre frontier node
>>>
>>>And additionally
>>>
>>> (depth == 3) //pre pre frontier node
>>>
>>>>(depth = 1) //frontier node
>>>>(depth = 0) //leaf node
>>>
>>>IMHO each node without any successor is a leaf node, so independent from depth.
>>>The opposite of leaf nodes are interior nodes.
>>>What about horizon-nodes for nodes at depth == 0?
>>>
>>>>(depth < 0) //qsearch node
>>>
>>>I would define a qsearch node for depth <= 0.
>>
>>
>>1)In that case all the leaf nodes are qsearch nodes.
>
>Only if you define leaf nodes exclusively for depth == 0.
>What about nodes with depth > 0, but stalemate, mate or repetition or even
>hashhit foreward pruning nodes? Aren't that leaves or terminal nodes in the
>current search space as well?
You are right.
I thought that the leaf nodes are subset of nodes when depth<=0 because in most
cases when depth>0 I do not prune but you are right and there are cases when I
prune and it can be also not in case of stalemate or mate but in cases that I
used pruning because the evaluation of the side to move is bigger than beta
+margin(depth,...)
>
>IMHO leaf nodes are the opposite of interior nodes, independent of depth
>less/equal/greater zero, only about having successor or not - may be i am wrong
>with this definition.
>
>
>>I think that it is not a logical definition because qsearch nodes mean for me
>>additional nodes that I have because of the qsearch and it should include only
>>nodes that are based on captures or checks that were done in the qsearch.
>>
>>2)For me there is depth of the qsearch and depth of the search and they have
>>different meaning so I do not use depth<0 to decide about qsearch nodes.
>
>
>My "definition" of depth == 0 nodes is horizon nodes, but not necessarily leaf
>nodes. Since qsearch handles such horizon nodes, i would consider qsearch nodes
>as nodes with depthleft <= 0.
>
>int search (int depth,...)
>{
> if ( depth <= 0 )
> return qsearch(...)
> ++nodeCount;
> ...
>}
>
>int qsearch (int depth,...)
>{
> ++qnodeCount;
> ...
>}
>
>Gerd
The problem is that by your definition qsearch nodes do not mean additional
nodes that are searched because you have qsearch instead of static exchange
evaluation.
In the opening position if you search to depth 1 all the nodes after 1 ply are
nodes with depth=0 so they are qsearch nodes by your definition but you will
search the same nodes even if you have no qsearch and only static exchange
evaluation that does the same job of the qsearch.
When I want to test the quality of my qsearch then one question is how many
additional nodes I search in the qsearch.
Uri
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.