> From: David Parter <[log in to unmask]>
> To: Multiple recipients of list
>
> > With all of this talk lately about RPICH and others & and how they
> > *all* seem to be so "unfair" in various respects, how difficult
> > *would it be for an
> > informed math/stats guru to develop a new method which takes all of the
> > meaningful criteria into account?
>
> In order to make a more "meaningful" or "accurate" rating, someone has
> to decide what *exactly* the criteria should be.  Right now, with
> a multitude of rating systems available, there is no consensus on
> what the correct ordering should be and why.
 
ABSOLUTELY NOT,  just look all the fun we are having.  Remember the
unscientific method is
1) collect a bunch of data
2) wonder what to do with the data
 
> > Also, if the existing methods aren't meaningful enough, why hasn't
> > someone in a position to change it taken any step in the right
> > direction?
>
> given that the selection committee still has some discretion, they seem
> reasonably satisfied with the current system. The biggest complaints
> that I have heard are not the details of the RPI, but using any kind of
> rating at all, instead of doing it purely subjectively.
 
I'm certainly not a proponent of the RPI.  I still firmly believe it
unjustly penalized a team last year (3 guesses on which team) and I
also believe the selection committee (last year) absolutely followed
the numbers without using any of the discretion david refers to
above.  Something like RPI might make more sense if we had lots of leagues
of vastly different quality.
 
charlie shub   University of Colorado at Colorado Springs
[log in to unmask]  -or-  cdash@colospgs (BITNET)
(719) 593 3492               (fax) 593-3369