Author: Roy Eassa
Date: 06:49:52 01/12/01
Go up one level in this thread
On January 11, 2001 at 17:35:08, Dann Corbit wrote: >On January 11, 2001 at 17:31:46, Garry Evans wrote: >[snip] >>>>Don't we have enough 40n 2 games to give a USCF type rating? >>> >>>Not even provisional. You can't combine every machine/program combination and >>>call it a single player. >> >> You are incorrect again, only 20 games need to be played to get a uscf >>established rating, Rebel during the GM challenges has played over 24 at 40/2 >>including the lithuatian challenges > >He has changed hardware several times. However, it is possible that enough >games have been performed by Rebel on some particular platform to get a >provisional rating. > >This (however) is a long, long way from proving how strong something is. On the >other, other hand -- I think that for the Rebel program we do have better >evidence than for any other. Doesn't this whole long-winded discussion come down simply to a definition of what it means to "prove" something? Using the strict mathematical/scientific definition of "proof", Mr. Corbit is correct. Nothing has been "proved". In the 1990s, in the US, O.J. Simpson was "proved" not guilty of a double murder in a criminal trial, and "proved" guilty of causing the same deaths in a civil trial. The burden of proof was different. The first trial required proof "beyond a reasonable doubt", whereas the second required only a "preponderance of evidence". Has it been proven mathematically that chess programs are GM strength? No. Beyond a reasonable doubt? That's debatable. By a preponderance of evidence? I'd say yes. (I haven't heard this kind of argument since my M.I.T. days!)
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.