Greg Weintraub writes:
>I know the validity of various ratings systems has been discussed at
>great length, especially the RPI because of its use in the NC$$ selection
>process. A frequent comment has been that ratings are meant to be used
>at the end of the season. Well, there are only 2 weekends left, so I
>think it's fair to begin analyzing the rating systems' performances.
I think it's still too early...more in a second.
>Looking at this weeks ratings is troubling. Colorado leads the WCHA by a
>substancial margin, yet trails Minnesota in the ratings. Colorado has a
>significantly higher win%, but Minnesota is ahead by virtue of its
>strength of schedule (beating UM and MSU during T-day certainly helped,
>as did playing BU over New Year's)
Well, first, the difference between CC and MN's RPI ratings is a
miniscule .0002. So, RPI considers the slightly higher Win% by CC to
cancel out with the much stronger strength of schedule that MN has.
I say CC's Win% is slightly higher because what matters in comparing
all the teams is the value of each team's Win% relative to the others.
CC's Win% is .7353 to MN's .6094, but when you rank the teams 1-44 by
Win%, CC is 4th and MN 10th. That's not much of a difference. On the
other hand, CC's OppWin% ranks 22nd, but MN's is 2nd.
There are also two other very important factors at work here. The
first is that the WCHA plays an unbalanced schedule. CC and MN each
had two teams they played only 2x as opposed to 7 teams they played
4x. But these were different for each school. Here are the teams
these two schools each played only 2x (team's Win% rank in parentheses):
CC Wisconsin (14), Minnesota-Duluth (28)
MN Alaska-Anchorage (36), Northern Michigan (34)
This means that another variance in schedules is that Minnesota played
Wisconsin (14) 4x and UMD (28) 4x, as opposed to CC which played those
teams only twice. And CC played UAA (36) and NMU (34) 4x each, while
Minnesota only played them twice each. That's a pretty big swing of
four games against stronger league opponents alone that Minnesota played.
The other important factor is why it is STILL too early to evaluate
the system. Each team has four games left:
CC Michigan Tech (21) 2x, Denver (7) 2x
MN North Dakota (26) 2x, St Cloud (27) 2x
Clearly, the strength of the teams' remaining regular season schedules
favors CC pretty heavily. But these games haven't been factored in yet.
>Similarly, it's hard for even me to justify BU being ahead of Maine (but
>who am I to argue :-) Maine has the best win% in the country, yet it's
>rated 3rd because of SOS.
Maine has the best Win%, but BU is right behind in 3rd. And here,
like with CC-MN, the difference is a tiny .0004, such that RPI
considers the difference in SOS to pretty much equal the difference in
Win%.
Since the difference is small enough, other factors would come into
play in comparing these two teams - factors like head-to-head, where
Maine would have a strong advantage of +2 (2-0-2). So even though the
RPIs may be equal, Maine may come out ahead in a full comparison. I
don't bother looking at this stuff until we get really close to
selection time.
>The most eggregious: NORTHEASTERN
>Here is a .5000 team rated 8th in the country, ahead of Clarkson with a
>.6667 win%. That's a HUGE difference. Also, compare NU to Lowell. They
>are 4 and 5 respectively in HE. Both are exactly .5000 overall.
> 8 NU
> 28 Lowell
>NU is a talented team that could be "ranked" highly in a subjective poll
>if they had started slow and since caught fire (much as LSSU has done).
>But a .5000 team should in no way be "rated" this high, especially
>compared to the closest team in its conference, Lowell.
What you seem to be saying here is that if teams rank close to each
other in their conference standings, they should also rank closely in
a rating. But the selection committee says this has nothing to do
with selection at all. What matters is a team's entire schedule, and
when you exclude the HE games, there is quite a bit of disparity
between who NU played and who UML played. I'll also toss in Merrimack
here for comparison, since Merrimack is 6th in HE (NU, UML, and MC are
all fighting for the last home ice spot) and just one game over .500,
but ranks 27th in RPI.
First, the three teams' numbers:
RPI Opp OppOpp
RK G W L T Win% RK Win% RK Win% RK RPI
8 Northeastern 31 13-13- 5 50.00 22 57.21 1 50.66 13 53.77
27 Merrimack 30 13-12- 5 51.67 20 47.43 38 49.56 27 49.02
28 Mass Lowell 34 15-15- 4 50.00 23 47.71 36 50.02 17 48.86
All three teams have played 21 HE games. Here are their nonconference
DivI opponents, with those opponents' Win% rank in parentheses. I
listed them from top to bottom, in decreasing order of Win% rank.
NORTHEASTERN - 10 MERRIMACK - 9 MASS LOWELL - 13
Boston University (3) RPI (15) Clarkson (6)
Denver (7) RPI (15) Brown (11)
Bowling Green (8) Miami (16) RPI (15)
Lake Superior (12) Colgate (18) Colgate (18)
Harvard (19) Alaska-Fairbanks (39) Colgate (18)
North Dakota (26) Alaska-Fairbanks (39) Western Michigan (24)
North Dakota (26) Air Force (40) St Lawrence (29)
St Lawrence (29) Air Force (40) Illinois-Chicago (32)
Yale (33) Army (44) Yale (33)
Dartmouth (38) Alaska-Fairbanks (39)
Air Force (40)
Air Force (40)
Army (44)
It is clear that NU had a much tougher set of NC games than either MC
or UML. BTW, remember the discussion before about how Win% is
actually worth about 65% of RPI rather than 25%...this is why
Merrimack's rating is barely higher than UML despite UML having a
slightly better strength of schedule.
There's another factor involved here, like in the CC-MN comparison...games
remaining. Northeastern still has to play UMass-Amherst THREE times,
while MC and UML have played UMA 3x already. This doesn't really deal
with the NU vs UML situation, but it does address the NU vs Clarkson
question.
I added in results for the last three regular season games for all
three of NU, MC, and UML, giving wins to the favorite in each case -
3 NU wins over UMA, 2 MC losses to BU and one win over BC, and 2 UML
losses to Maine and one win over PC.
MC and UML only moved up one spot each - of course, they each had a
win and two losses added in. NU dropped two to 10th even by sweeping
UMA 3-0-0. And, NU's RPI dropped low enough that they were REALLY
anywhere from 7-13, as opposed to 7-11. That means that on their own
merits, NU could very well fall back out of contention depending on
how the other factors stack up for them. They could end up being the
victim of the 2 teams per conference rule just as easily as Wisconsin
or BG could.
And of course, none of this factors in playoff results or results of
other games the rest of the way.
>It's hard to
>accept a team such as BGSU (currently 12) getting bumped out of the
>tournament in favor of having all conferences with at least 2 bids (Brown
>being rated 13 and getting the 2nd ECAC bid). It's even harder to think
>of Wisconsin or Denver losing a spot in favor of a 4th place HE team with
>a .5000 record. Granted, Wisconsin is currently 4th in the WCHA, but
>they've got a better record and are still in contention.
If NU sweeps UMA as they'd be expected to do, their record would be
16-13-5. Wisconsin is 16-12-4 right now with games left against
Denver 2x and MTU 2x. It's very possible the two teams will have
about the same Win% by the end of the regular season - and their SOS
will be similar, too. A case could certainly be made for NU over
Wisconsin at that point.
>After getting
>swept by UNH, NU is in jeopardy of losing home ice.
But not seriously, since they have by far the easiest remaining
schedule of the three teams in contention for that final home ice
spot, and they are also the team that the other two have to catch.
And, they win the tiebreaker over both UML and MC.
>I'm all for having 4 HE teams in the tourney, but NU just hasn't earned
>it (yet, a strong finish plus a strong showing in the HE tourney could
>change that). NU has benefited from a strong NC schedule (BU and
>Harvard in the 'Pot, Badger Classic, a game with DU) in which it
>has performed poorly (if memory serves me a 3rd, a 4th and an embarassing
>home loss respectively).
Yes, that strong NC schedule is what has helped them. But a poor
performance? Not in the least. NU went 5-5-0 over one of the tougher
NC schedules in the country. BTW, RPI doesn't care who you beat and who
you lose to, so NU's losses in the tough Badger Showdown could have
been changed to wins and their wins over Dartmouth and St Lawrence
changed to losses, with no change in ranking at all. (I checked it.)
At any rate, that loss to Denver that originally seemed so embarrassing
sure isn't looked upon that way now...Denver went on to prove itself
as one of the top teams in the country this season.
Like many, I happen to believe Northeastern isn't playing up to what I
expected from them before the season, but they still deserve the same
look as everyone else when it comes to handing out bids. And by going
(currently) .500 against a tough schedule, they've still made a strong
case for themselves. Not many teams have distinguished themselves
this season.
--- ---
Mike Machnik [log in to unmask]
Cabletron Systems, Inc. *HMM* 11/13/93
|
