Deron Treadwell writes:
(I wrote):
>>Where this is important is in Maine's race with BU for the #1 seed in
>>the East. The loss makes the difference between BU and Maine 1.37 as
>>opposed to 0.60. Since the threshold below which more detailed
>>comparisons can occur (including head to head record) is 1.00 (0.01 in
>>RPICH), a difference of 1.37 means that Maine's +2 head to head
>>advantage (2-0-2) over BU would not come into play if the season ended
>>today. BU would be the top seed in the East.
>
>This says something about the validity in the system the NC$$ uses then. If
>a team wins the division (assuming the season ended today Maine would win
>Hockey East) has a higher winning % and wins the head to head convincingly
>(2-0-2) and finishes behind the other team in the NC$$ tournament rankings,
>something is messed up.
I may duplicate some of what Wayne said in his reply, but I hope there
will be sufficient new information here so as to make this reply useful.
Wayne mentioned that the RPI doesn't care about conference finish. This
is because the intent is to select teams based on their performance
throughout the entire season, not just within the conference. The RPI
doesn't even know or care that conferences exist.
As such, it makes sense (to me anyway) that a team which finishes
below another in the conference could rank higher in the RPI. It goes
back to the value of nonconference games played, and HE teams play 10
of their (approximately) 34 games outside the conference. Who you
play and how you fare in those 10 games could be enough to lift you
ahead of another team.
Where conference finish matters, is in that the selection committee
has decided to select the regular season conference winner for a bid
even if that team does not earn the conference's automatic bid by
winning the conference tournament. Other than that, forget it. This
is another common misconception: every year, you will hear people
complain about the fact that Team B finished 4th in their conference
and Team A finished 3rd, but Team B got a bid and Team A did not.
Doesn't matter.
>I think we know this already since Colorado College
>was so "quaintly" excused from the tournament last year.
However, as an analysis last season showed, CC was not kept out simply
because of the RPI. CC was in a battle with several other teams for
the last few spots. The battle was so tight that the teams' RPI
rating fell within the .01 threshold, so that other factors had to be
looked at - head-to-head, Last20, CommonOpp, etc. CC did not measure up
in those factors to the teams with which it was in contention. It is
true that a better RPI would have gotten CC a bid, but so would a
better performance in the other factors.
>I know that the
>season has a while to go yet, but Mike does bring up an excellent point
>here. Maine simply by playing UMA lost points in the power ratings, the
>loss is magnified by the defeat.
Actually, this is not true! I did another analysis where I changed
the Maine-UMA game back to a win for UMA, and instead, I made Maine's
loss to Princeton a win. Maine's rating rose as it did when the UMA
loss was changed to a win, but it is interesting to note that even
though Princeton is rated much higher than UMA, the difference between
BU and Maine was *greater* than it would have been if the win had come
over UMA. Not by much (.0005, I believe), but still greater.
And note that this is where the RPI numbers really matter. The actual
figures aren't important. What counts is your rating relative to the
rating of other teams.
Another important point is that Maine was bound to take a hit just
from playing UMA twice. This is because as Erik has said to me before
(maybe to the list too), the RPI is really not meant to be looked at
during the season. It only matters at the very end after all games
have been played. Before that, discrepancies are bound to exist
because, for example, teams will not have played the same number of
games against other conference teams. BU had already played UMA twice
before Maine even played them once. Thus, prior to last weekend, BU's
OWP was really "lower" relative to Maine's than it was bound to wind
up at the end of the season.
You will see BU take a hit as well when BU plays UMA Friday, because
then BU will have played UMA 3x to Maine's 2x.
>Here's an opinion, does anyone else think that everything, power-ratings,
>winning percentage, division finish and head to head competition should ALL
>be taken into effect by the NCAA? I realize winning % is included in the
>power-ratings, but what about these other factors? I'll certainly be
>interested to here your opinions!
In one way or another, all of these factors do come into play in
seeding the teams. Whether they should (and how) is going to be
something not many people will agree on. There are some legitimate
concerns people do have with both the way RPI is calculated and the
way the selection process plays out. But as Wayne said, we can take
solace in the fact that the process has evolved to the point that
there is a minimal amount of bias involved in the selection. In that
regard, I have been very pleased with the way selection has been
handled in recent years.
My chief bone of contention with the process is that even when two
teams have their rating fall within the .01 threshold, and then the
additional factors are looked at (Last20, CommOpp, HeadToHead, etc.),
it makes no sense to me that RPI rating should also be one of these
factors. The fact that a comparison proceeds to these other factors
indicates to me that the committee agrees that the closeness of the
rating makes it too close to call on RPI alone. But then if the
comparison by other factors produces a tie, the tiebreaker is RPI. I
would like to see a different tiebreaker used.
Just to clarify, the process is (roughly):
1) Use RPI to rank teams. If any two teams have an RPI within .01 of
each other, proceed to step 2 in comparing those two teams.
2) Compare teams using several factors, including RPI, HeadToHead,
CommOpp, Last20, & RecordVsTeamsUnderConsideration. Award one point
for an edge in any factor, except that for HeadToHead, award points
equal to the difference in HeadToHead - for example, a record of 3-1-0
results in +2 for the team with 3 wins. If teams are tied, proceed to
step 3.
3) Break ties with RPI.
I am unsure off the top of my head what happens if in step 3, the RPIs
somehow worked out to be exactly equal. I don't believe that it has
ever gone beyond that.
I am also unsure what happens in the following scenario:
A, B, & C all have an RPI within .01 of each other. In the ensuing
comparisons, A ranks ahead of B, B ranks ahead of C, and C ranks ahead
of A.
If anyone can answer this, I know I would greatly appreciate it.
Otherwise, it is something for me to ask the People Who Know, I guess.
--- ---
Mike Machnik [log in to unmask]
Cabletron Systems, Inc. *HMM* 11/13/93
|
