Author: Jasmine Baer
Date: 12:30:14 01/02/04
Go up one level in this thread
On January 02, 2004 at 10:46:25, Luis Smith wrote: >On January 02, 2004 at 10:42:37, Jasmine Baer wrote: > >>On January 02, 2004 at 10:35:40, Luis Smith wrote: >> >>>On January 02, 2004 at 10:31:23, Jasmine Baer wrote: >>> >>>>On January 02, 2004 at 10:21:48, Luis Smith wrote: >>>> >>>>>On January 02, 2004 at 09:57:38, Jasmine Baer wrote: >>>>> >>>>>>I've seen it written that under the following conditions: >>>>>> >>>>>>1. Engine vs. Engine match or tournament >>>>>>2. Held on a single computer with a single processor >>>>>> >>>>>>having ponder=ON(or Permanent Brain in the Fritz GUI) will impact the play of >>>>>>the engines since the each individual engine would not have full access to the >>>>>>processor during its own turn. >>>>>> >>>>>>First, is this true? >>>>>> >>>>>>Second, is this issue, if it actually is an issue, something that is eliminated >>>>>>by running a two-processor system? >>>>>> >>>>>>And, finally, does anyone have any solid insight on how ponder=off/on or >>>>>>Permanent Brain works on a Pentium 4 with Hyperthreading? >>>>>> >>>>>>Thanks. >>>>> >>>>>No need, someone has already run this test. Click on "Winboard" and then >>>>>"Ponder Experiment" on the side >>>>> >>>>>http://www.pittlik.de/winboard.html >>>> >>>>Am I reading the time control corrctly? 60+2 meaning 60 minutes plus 2 seconds >>>>per move? The reason I ask is because above the tables, there are links to some >>>>downloads - 1320 Lightning Games etc., etc. 60 minutes + 2 sec/move is not a >>>>lightning time control. Could I be misinterpreting the 60+2? Should I be >>>>thinking that 60+2 means 60 seconds plus 2 seconds per move? >>> >>>60 seconds with 2 second increments. >> >>Then it makes sense that ponder on/off doesn't make any statistical difference. >>The computer hardly has time to think under that time control. I'm guessing >>that the longer the time control, the more influence ponder on/off has. > >How can you say that it does not make any statistical difference? 60/2 is a >very unstable time control, so if he gets results like I think it says >something. > >Although you might be correct that longer time controls could effect the outcome >more, I am getting a little bit more interested in your experiment, however I >don't have a dual or 2 machines to test this one. "The results of both tournaments are very similar. The differences are within the usual statistical error margins. Tournaments with ponder=off and ponder=on seem to give comparable results." This quote is from the link included in a previous response (http://www.pittlik.de/winboard.html). I should have been clearer. INstead of saying it doesn't make any statistical difference, I should have said that I am nos surprised that the differences in performance between ponder=on and ponder=off fell within the statistical margin for error.
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.