Not often I get to disagree with Mike, but here goes ...  :-)
(only portions of Mike Machnik's article shown here)...
 
> The RPI is not meant to be looked at midway through the season.  It has one
> sole purpose: it is to be used in selecting the teams for the NCAA
> tournament after all the games are played.  After ALL games are played.  It
> is not intended to be used at a random point in the season to tell us
> anything about the teams.
 
Here's my disagreement.  The NCAA uses RPI in (a large part of) the
selection process for the end of season tournament, but I don't think
it follows that RPI has "one sole purpose".  Anyone with interest in
nationally ranking college hockey teams through the season should use
the information available ... and I believe RPI contains information to
that end.  So RPI now is not interesting to the NCAA, but may be to us.
 Let me emphasize this with two points.
 
RPI *will* be used at end of season in the NCAA ranking process.  It
makes sense for us to see and understand RPI during the period leading
up to the end of season, IMHO.
 
There is nothing magical about the "end of season", except that the
NCAA will make it's determination then.  Schedules will *not* be
symmetric then, because college hockey schedules are not symmetric.
League schedules will be complete, but they are not symmetric in many
cases.  League tournaments can have a large effect on RPI rank;  ask a
certain team of a couple of years ago what it's like to win the
conference and not be invited to the NCAAs, in large part because of a
first round loss in the conference tournament to a "weak" team.
 
So we see anomalies in current RPI rankings, but we should not say
those are unimportant ... the effects, diluted somewhat, will be with
us at end of season.  And those anomalies may help us to understand
some teams better.  CC seems to be able to blow away "weaker"
opposition, but will then play consistently well against the upper
echilon?  Probably, but it remains to be seen!
 
>                                               ...   A better analysis would
> be to wait until the end of the season and run the RPI twice: once with all
> games, and again without the CC-AF game.  I believe that the effect of that
> game will be minor if any at that point.
 
Yet the decisions of placement or entry into the NCAA tournament are
made with data that are so close that each game is not minor, but
important in some cases.
 
> Perhaps now we can see why the NCAA Selection Committee does not wish to
> release the RPI during the season.
 
Each year it seems that the NCAA will provide a periodic release ...
only to back down when the time comes.  There was some talk of the NCAA
providing a periodic (monthly?) RPI, to start in January of this
season, I believe. We'll see.
 
> But I do think it is a good question to raise, that a team can drop by
> winning a game against a weak team.  It doesn't make sense if you believe a
> team should never drop when it wins.  It may make sense if you believe that
> strength of schedule should be a component in the rating.  People will have
> to form their own opinions on this.
 
Right on!
 
Regards,
 
Wayne T. Smith
Systems Group -- CAPS        internet: [log in to unmask]
University of Maine System   BITNET/CREN: WTS@MAINE
 
HOCKEY-L is for discussion of college ice hockey;  send information to
[log in to unmask], The College Hockey Information List.