Dick Tuthill writes:
 
>        The fact that the fluctuations in ratings damp out with time over the
>course of a season has nothing to do with the efficacy of the various
>methods.   It is simply an intrinsic property of the averaging process as
>more results are added to the averages.   It is a misinterpretation of our
>criticism of the RPI to say that ... "yeah,  but all this bouncing around
>in the ratings will settle out in the end."   That's not the point.
>
>        Sure,  we know that the RPI individual ratings will converge to
>stable values.   But,  will the RPI give "equitable" converged results??
>Any system which rewards a team for going to Maine and losing two is not
>trying to measure accomplishment.   What it is doing is simply social
>engineering.   And,  as I've pointed out before,  that may be needed in
>basketball but it isn't in college hockey.
 
I think a significant point is being missed here.  The only purpose for
calculating the RPI is to select the top 12 teams for the tournament.  The
RPI is applied once, after all games preceding the NC$$ tournament have been
played.  It's not meant (I believe) to be a comprehensive weekly rating of
all the Div1 teams.  Week-to-week changes in a team's rating aren't
significant.  All that matters is where the teams stand when the final RPI
is calculated.  RPICH just gives us a way of making informed guesses at
weekly intervals about the selection process.  Perhaps we're discovering why
the committee doesn't release RPI numbers to the public...
 
-- Erik