Author: Ed Schröder
Date: 10:11:12 12/30/01
Go up one level in this thread
On December 30, 2001 at 12:55:07, Ed Schröder wrote:
>On December 30, 2001 at 10:57:30, Severi Salminen wrote:
>
>>I just would like to see how other engines do in similar position. So, feed the
>>position (1.e4 e5 2.d4 d5) to your engine, perform a 10 ply search and report
>>the number of nodes and qnodes. To make sure we count nodes identically, this is
>>what I do: I increase nodes if I don't call qsearch() from search() and I
>>increase qnodes in the beginning of qsearch(). Like this:
>>
>>search()
>>{
>> if(depth<=0)
>> qsearch()
>> nodes++;
>>.
>>.
>>}
>>
>>qsearch()
>>{
>> qnodes++;
>>.
>>.
>>}
>>
>>Try to get similar counting scheme so figures are comparable. Here is the
>>position:
>>[d] rnbqkbnr/ppp2ppp/8/3pp3/3PP3/8/PPP2PPP/RNBQKBNR w KQkq - 0 1
>>
>>And please tell if you use SEE or not. I'll post my results soon also.
>>
>>Severi
>
>
>Rebel: Total nodes: 4.580.488
> QS nodes : 747.568
>
>I don't know about SEE, when Rebel evaluates a position I get back the 3 highest
>hanging pieces (possible losses) and the 2 highest attacked pieces (threats), so
>I don't need SEE, it is build-in the evaluation so to say.
Update:
Total nodes: 4.580.488
QS nodes : 747.568
When I remove checking moves from QS, so a pure QS, I get:
Total nodes: 3.346.349
QS nodes : 343.423
Just under 10% :)
Ed
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.