Author: Roy Eassa
Date: 14:07:09 12/06/00
Go up one level in this thread
Standard statistical science says that the smaller the sample, the greater the chance that the results are due to random chance. There are formulas to calculate the exact probability of random chance, given the sample size (although I don't know the formulas). That's why you see something like Gore 46%, Bush 42%, with the margin of error being specified as, e.g., "plus or minus 3 percentage points". Or why a DNA match will be with 99% or 99.9% or 99.99% or 99.999% certainty. On December 06, 2000 at 07:17:54, Francesco Di Tolla wrote: >On December 05, 2000 at 17:59:55, Timothy J. Frohlick wrote: > >>Chris, >> >>In other words, you are saying that these two programs are about equal in >>strength. I accept that. The folks that keep saying that "we can not know the >>real strength of a program until we have played at least N-hundred games" are >>being picayune. I know that a 500 Mile/hour rocket sled is faster than a >>Volkswagen just after one run. I think that you can get a good picture of a >>programs strength after twenty to fifty games. >> >>Both Shredder 5.0 and ChessTiger 13 are exceptional programs. >> >> >>Tim > >In other word he is saying: I can't tell the diffeence. Which is quite different >from what you say. A small test with little or no difference says we cannot see >it. The bigger the test the smaller the minimal difference you can resolve. >But this is not saying that the two program play the same level. > >Pushing your conclusions ad-absurdum, a signle draw between any two >pograms/players would imply they are at the same level. > >When you measure something you should estimate the uncertainity of the measure, >otherwise you are measuring absolutely nothing. > >regards >Franz
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.