Author: Don Dailey
Date: 08:44:20 06/24/98
Go up one level in this thread
On June 24, 1998 at 09:47:15, Robert Hyatt wrote: >On June 24, 1998 at 04:08:23, Inmann Werner wrote: > >>What are nodes per second. >>I count, how often alpha-beta is called in a search, and taht are my nodes. >>Is this correct, or should I count the generated moves? >>It is not really important, but everybody speaks about, how many nps a program >>makes, and if everybody defines nodes different, all compares a nonsense. >>On the other hand, nps only give sense, if there is no computer thinking on >>opponents time!? >> >>Werner > > > >that's what most of us are doing... and your count should be >comparable to ours... just add nodes++; to the top of search() >and quiesce(); Technically (or sematically) this is correct, but you need to be careful. I count how many moves are actually executed in a search (or in my program calls to make_move(). I think this is more accurate. Depending on how your program is structured counting calls to alpha-beta could give you the wrong answer. An early program of mine would sometimes do a static evaluation without making this final call, using your method would have seriously deflated the node counts. Really though this is an issue of semantics. But a node is a new position visited and not an issue of how you structure your recursion. I would guess that by far most programs this would work the same. - Don
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.