Author: Robert Hyatt
Date: 07:34:01 10/08/02
Go up one level in this thread
On October 08, 2002 at 03:01:00, Kurt Utzinger wrote: >On October 07, 2002 at 23:11:32, Robert Hyatt wrote: > >>On October 07, 2002 at 18:33:43, Jeremiah Penery wrote: >> >>>On October 07, 2002 at 18:29:35, Uri Blass wrote: >>> >>>>On October 07, 2002 at 18:19:29, Jorge wrote: >>>> >>>>>This may be a dumb question, but could someone tell me the reason for setting >>>>>Ponder=off in comp to comp matches? If two computers with the same hardware with >>>>>different, each with a program playing each other, why set ponder off? >>>>> >>>>>thnx, >>>>>jorge >>>> >>>>Usually the people who play games with ponder off use only one computer and not >>>>2 computers. >>> >>>Why set ponder off with one computer? >> >> >>The basic idea is that one computer matches are not very informative. You have >>two >>choices: >> >>1. Ponder=on. Engines can "steal" time from each other since both will be >>computing most >>of the time. Unless one is unlucky enough to not have anything to ponder. For >>that move, the >>opponent runs twice as fast. >> >>2. Ponder=off. This might cause poor time utilization as it is an unnatural >>way to run and >>is not always very well tested. >> >>Either way causes problems. Moral? Use two computers or else don't pay a >>significant amount >>of attention to the results. > > >Previous and still running tests with quite a lot of games in Germany prove that >there is no significant difference in results between matches played with either >ponder on/off. And most surprisingly - contrary to your assumption - this also >refers to Crafty. >Kurt So? You are aware that I changed the time allocation in Crafty a year or two back to solve this specific issue? Prior to that it _definitely_ hurt. Since I never tested that way, I never tuned the timing code at all. I suspect there are others that did that as well. Which means the results are _always_ subject to doubt, IMHO. Just like racing a couple of cars downhill with the motors off. Which one wins? Will it be the fastest overall with engines _on_??? Maybe or maybe not... To see, you have to run _both_ tests. By then, it doesn't matter as you have run the _right_ test... Same for chess engines... > > >The running test
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.