Dick Tuthill writes:
>   First of all,  why is the subject important??   Because the RPI
>is used by the NC$$ in the tournament selection process.   And,   ...
>last year CC got royally screwed by it  --  to put it politely.
 
Well, I don't feel this was the case - I think there were legitimate
reasons why CC did not make the final 12, most of which were under
CC's control and which I discussed in great detail last spring (hmm,
and into the summer too :-)).
 
I think that before I buy into a new method which is meant to solve
whatever problem was supposed to have existed, I'll need to be
convinced that there was a problem in the first place.
 
However, let me point out that whereas the NC$$ Selection Committee
apparently did feel that there was a problem (or else felt enough
pressure to change things), they changed the process this year so that
they will now take the tournament AND regular season champions from
each conference.
 
So I also have to ask, since a situation like CC's could never happen
again (unless the rules are changed), why is it necessary to make any
change in the RPI at all, even if I do agree that CC should have
received a bid last year?  What are we trying to prevent?
 
>And,  I would submit that the form which I propose weights win% equally
>with strength of schedule.
 
What says that this should be the case?  I'm not disagreeing, but I do
want to hear why this is correct before I say I agree or disagree.
 
>   Finally,  the fatal flaw in RPI is that a team can play a tough
>schedule,  win *NO* games,  and yet get a pretty good rating simply because
>strength of schedule is 75% weight and *additive* to win%.
 
I don't believe this is true at all.  I would have to see proof of
this before I agree...because when I look at past RPIs, I see a
boatload of teams with a poor Win% and strong OWP & OOWP, and they are
all ranked near the bottom of the RPI.  I don't see any team in this
category that is anywhere close to receiving a bid, and the only teams
close are ones who had a sub-.500 record but were still much better
than .000.
 
I feel very confident in saying that no team could have this type of
a season (win 0 or few games, play strong schedule) and still be able
to receive a bid.  For one thing, I don't think RPI is perfect, but it
is not that bad.  And for another, take the example of Maine last
season.
 
Maine's schedule strength was very strong, but it ranked about 17th in
RPICH because of its 14-18-4 record (not including the Tardif forfeits).
(Imagine how much lower Maine, with just about the strongest schedule
in the nation, would have ranked with a Win% of .000.)  Still, because
of the .01 threshold, Maine could have been seeded as high as 11th,
which would have been good enough for a bid.
 
But then the other factors came into play, factors which were posted
by JohnH a little while ago from the CCHA press release.  And here,
Maine stood no chance against any of the other teams it was in
contention with.  That's because its weaker record came into play in
creating its weaker Last20Win%, CommonOppWin%, etc.
 
No team with a .000 Win% or even a very poor Win% would ever win out
against other teams in head to comparisons because of the combination
of the RPI and the additional factors that come into play.  So, to me,
there is no problem here that needs to be corrected.
 
>A good example
>of this sort of thing (although less dramatic) is Harvard's ranking in
>the latest RPI.   I'm sorry,  but I have a REAL problem with a 0.500 team
>in the national top ten.
 
I'm sure Erik's response is going to be that this is why he didn't really
want to begin posting the RPICH right now as it is, but he decided to do
it anyway for everyone's benefit.  There have *still* been too few games
played for it to mean much.  So take any RPI you see now with a huge
grain of salt.
 
Cornell was 13th in a Nov 27 edition Erik sent me, with a record of
1-2-1.  Good thing he didn't post it then!  There really would have
been an uproar. :-)
 
>I think 14th place (just out of the tournament)
>is a much more appropriate ranking for Harvard at this point in time.
 
Why 14th?  Why not 11th, or 13th, or 20th?  To me, it seems just to be
a value judgment based on the feeling that .500 teams aren't supposed
to be in the top ten and that 14th "seems" right.  That isn't enough
of a basis for me to agree that there is a problem with the RPI that
is solved by some other method.  If I tell you that I do think a .500
team could belong in the top 10 or 12 under certain circumstances
(i.e. playing a much tougher schedule than a team with a slightly
better and above-.500 record), how would you counter that?  This is
really what I am asking.
 
Again, I'm just trying to show that more evidence of a problem needs
to be shown, as well as good, solid evidence that any solution really
does work.  And we also need to remember that these aberrations in RPI
do tend to get taken care of as the season goes on and more games are
played.
---                                                                   ---
Mike Machnik                                            [log in to unmask]
Cabletron Systems, Inc.                                    *HMM* 11/13/93