It was recently pointed out to me that although we didn't know
it at the time, the NCAA has been using basically the same procedure
to seed the tournament for the past three years, since the switch from
RPI to pairwise comparisons took effect. So let's recap how the
decisions were made each time.
1996:
Six Western teams (2 WCHA, 4 CCHA), Six Eastern teams (3 ECAC, 3 HE)
Western Qualifiers
Team lPWR RPI Comps Won
1 CO College (W) 5 .615 MiLSMnWMMS
2 Michigan (C) 4 .615 LSMnWMMS
3 Lake Superior (C) 3 .605 __ MnWMMS
4 Minnesota (W) 2 .603 ____ WMMS
5 Western Mich (C) 1 .582 ______ MS
6 Mich State (C) 0 .582 ________
The comparisons would say to send WMU and MSU East, but
Michigan State is the host school, and must stay in their own region,
which would seem to indicate sending Minnesota and Western Michigan.
But that would have left three CCHA teams in the West regional and
neccessitated a potential second-round intraconference game, so Lake
State and Western Michigan were sent East instead.
Eastern Qualifiers
Team lPWR RPI Comps Won
1 Boston Univ (H) 5 .621 VtCkMLCrPv
2 Vermont (E) 4 .581 CkMLCrPv
3 Clarkson (E) 3 .576 __ MLCrPv
4 Mass-Lowell (H) 2 .569 ____ CrPv
5 Cornell (E) 1 .558 ______ Pv
6 Providence (H) 0 .549 ________
Okay, here's where things get tricky. The numbers say to send
Cornell and Providence West. The committee did send Providence, but
decided Cornell should stay East on attendance grounds. Lowell was
next in line to be shipped (which we didn't realize at the time, since
they happened to have a higher total PWR than Clarkson), but that left
all three ECAC teams in the East, leading to a potential second round
matchup. If the committee had been concerned only with avoiding those
matchups, they would have kept Lowell in the East, but they gave
preference to Cornell based on attendance and Clarkson based on
comparisons. I still don't understand the statement that Lowell was
shipped to avoid a potential second-round game with BU. Consider the
three possible regionals that would have resulted from sending
Cornell, Lowell, or Clarkson West along with Providence:
a) Cornell:
Team lPWR RPI Comps | Team lPWR RPI Comps
1 CO College (W) 1 .615 Mi | 1 Boston Univ (H) 1 .621 Vt
2 Michigan (C) 0 .615 | 2 Vermont (E) 0 .581
3 Minnesota (W) 3 .603 MSCrPv | 3 Lake Superior (C) 3 .605 WMCkML
4 Mich State (C) 2 .582 CrPv | 4 Western Mich (C) 1 .582 Ck__
5 Cornell (E) 1 .558 __ Pv | 5 Clarkson (E) 1 .576 __ ML
6 Providence (H) 0 .549 ____ | 6 Mass-Lowell (H) 1 .569 WM__
The natural brackets give no intraconference regionals games:
5W Cornell (E) 6E Mass-Lowell (H)
4W Mich State (C) 3E Lake Superior (C)
1W CO College (W) --+--2E Vermont (E)
|
2W Michigan (C) --+--1E Boston Univ (H)
3W Minnesota (W) 4E Western Mich (C)
6W Providence (H) 5E Clarkson (E)
b) Lowell:
Team lPWR RPI Comps | Team lPWR RPI Comps
1 CO College (W) 1 .615 Mi | 1 Boston Univ (H) 1 .621 Vt
2 Michigan (C) 0 .615 | 2 Vermont (E) 0 .581
3 Minnesota (W) 3 .603 MLMSPv | 3 Lake Superior (C) 3 .605 WMCkCr
4 Mass-Lowell (H) 2 .569 MSPv | 4 Western Mich (C) 2 .582 CkCr
5 Mich State (C) 1 .582 __ Pv | 5 Clarkson (E) 1 .576 __ Cr
6 Providence (H) 0 .549 ____ | 6 Cornell (E) 0 .558 ____
This was the option the NCAA chose. The natural brackets gave
one potential ECAC matchup, but with all three ECAC teams in one
regional, that's unavoidable; at least it required an upset (which as
we know didn't happen):
5W Mich State (C) 6E Cornell (E)
4W Mass-Lowell (H) 3E Lake Superior (C)
1W CO College (W) --+--2E Vermont (E)
|
2W Michigan (C) --+--1E Boston Univ (H)
3W Minnesota (W) 4E Western Mich (C)
6W Providence (H) 5E Clarkson (E)
(It was also a game between teams who tied twice in the
regular season and did not meet in the conference playoffs. The
explanation of this year's seedings seems to indicate this is
considered a less undesirable intraconference game.)
c) Clarkson:
Team lPWR RPI Comps | Team lPWR RPI Comps
1 CO College (W) 1 .615 Mi | 1 Boston Univ (H) 1 .621 Vt
2 Michigan (C) 0 .615 | 2 Vermont (E) 0 .581
3 Minnesota (W) 3 .603 MSCkPv | 3 Lake Superior (C) 3 .605 MLWMCr
4 Mich State (C) 2 .582 CkPv | 4 Mass-Lowell (H) 2 .569 WMCr
5 Clarkson (E) 1 .576 __ Pv | 5 Western Mich (C) 1 .582 __ Cr
6 Providence (H) 0 .549 ____ | 6 Cornell (E) 0 .558 ____
This gives two possible second-round conference matchups
(Cornell-UVM and Lowell-BU) under the natural seeding, but if you swap
LSSU with UML and WMU with Cornell (thus preserving the first-round
pairings) you have no intraconference games in either regional:
5W Clarkson (E) 6E Western Mich (C)
4W Mich State (C) 3E Mass-Lowell (H)
1W CO College (W) --+--2E Vermont (E)
|
2W Michigan (C) --+--1E Boston Univ (H)
3W Minnesota (W) 4E Lake Superior (C)
6W Providence (H) 5E Cornell (E)
The fact that they didn't do this is one reason I was surprised by the
idea of re-arranging seeds within regions in 1997.
1997:
Seven Western teams (4 WCHA, 3 CCHA), Five Eastern teams (3 ECAC, 2 HE)
Western Qualifiers
Team lPWR RPI Comps Won
1 Michigan (C) 6 .628 NDMnDUMmMSCC
2 North Dakota (W) 5 .588 MnDUMmMSCC
3 Minnesota (W) 4 .581 __ DUMmMSCC
4 Denver U (W) 2 .554 ____ MmMS__
5 Miami (C) 2 .577 ______ MSCC
6 Mich State (C) 1 .547 ________ CC
7 CO College (W) 1 .560 ____DU____
Eastern Qualifiers
Team lPWR RPI Comps Won
1 Clarkson (E) 4 .600 VtBUNHCr
2 Vermont (E) 2 .579 BUNH__
3 Boston Univ (H) 2 .577 __ NHCr
4 New Hampshire (H) 1 .591 ____ Cr
5 Cornell (E) 1 .570 Vt____
Okay, before we do anything, we have to take into account two
additional factors. First, automatic byes were given to the three
teams that won the regular season and tournament championships in
their conferences; Michigan and NoDak were already in line for those,
but BU gets the bye only for that reason. (I only just now realized
that BU's automatic bye was actually snatched away from Vermont, not
New Hampshire; UNH had a higher PWR because they won the comparison
with Maine, who were ineligible, but UVM would actually have gotten
the second bye based on comparisons.)
The second issue is the imbalance between the regions. Seven Western
teams means that one intraconference matchup in the West Regional is
inevitable as long as four of the five Eastern teams stay in the East.
Keep that in mind later. As a more immediate concern, we have to push
one of the Western teams into the East to start with. Now, this turns
out to be a little weirder than I had realized. According to Adam
Wodon's interview with Joe Marsh, this was the lowest-seeded team,
Michigan State. However, looking at the comparisons above, I would
have though Colorado College was the lowest-seeded team, since they
lose the individual comparison with Michigan State. This actually has
an interesting impact. Consider the starting regions if we dub MSU or
CC an Eastern team:
a) Michigan State:
Team lPWR RPI Comps | Team lPWR RPI Comps
1 Michigan (C) 1 .628 ND | 1 Clarkson (E) 1 .600 BU
2 North Dakota (W) 0 .588 | 2 Boston Univ (H) 0 .577
3 Minnesota (W) 3 .581 MmCCDU | 3 New Hampshire (H) 2 .591 __CrMS
4 Miami (C) 1 .577 CC__ | 4 Vermont (E) 2 .579 NH __MS
5 CO College (W) 1 .560 __ DU | 5 Cornell (E) 2 .570 __Vt MS
6 Denver U (W) 1 .554 Mm__ | 6 Mich State (C) 0 .547 ______
When the committee goes to swap the bottom two teams in each
region, it finds three-way ties which are not resolved by individual
comparisons. In each case they break the tie based on Ratings
Percentage Index and thus swap CC and DU with Cornell and MSU.
b) Colorado College:
Team lPWR RPI Comps | Team lPWR RPI Comps
1 Michigan (C) 1 .628 ND | 1 Clarkson (E) 1 .600 BU
2 North Dakota (W) 0 .588 | 2 Boston Univ (H) 0 .577
3 Minnesota (W) 3 .581 DUMmMS | 3 New Hampshire (H) 2 .591 __CrCC
4 Denver U (W) 2 .554 MmMS | 4 Vermont (E) 2 .579 NH __CC
5 Miami (C) 1 .577 __ MS | 5 Cornell (E) 2 .570 __Vt CC
6 Mich State (C) 0 .547 ____ | 6 CO College (W) 0 .560 ______
Now you see that, with Colorado College out of the mix, Denver
is ranked above Miami, which means a strictly by-the-numbers approach
sends Cornell and CC West and Miami and MSU East. Of course, this
would have put all four WCHA teams in the West, which would have set
up a first-round game between two of them. But still, starting with
CC in the East would have been pretty likely to send Miami East
instead of one of the Colorado teams.
What I think happened is that the committee decided to start
with MSU as the nominal East team, knowing they wanted them in the
West ultimately, and thus went with regions (a) to start with. The
other posssibility is that they broke the tie between the 6th and 7th
West teams based on RPI rather than the individual comparison, but
this would seem to go against the usual procedure of looking at the
individual comparisons. Ultimately, it doesn't matter; the committee
has discretion to choose which two Western teams go East, and they
decided it would be the two Colorado teams (positions 4-7 in the West
were plagued by non-transitive comparisons anyway). The regions after
swapping were thus:
EAST | WEST
1 Michigan (C) 1 .628 ND | 1 Clarkson (E) 1 .600 BU
2 North Dakota (W) 0 .588 | 2 Boston Univ (H) 0 .577
3 Minnesota (W) 3 .581 MmCrMS | 3 Vermont (E) 3 .579 NHCCDU
4 Miami (C) 2 .577 CrMS | 4 New Hampshire (H) 2 .591 CCDU
5 Cornell (E) 1 .570 __ MS | 5 CO College (W) 1 .560 __ DU
6 Mich State (C) 0 .547 ____ | 6 Denver U (W) 0 .554 ____
Then, in a move very unpopular in Minneapolis but obvious given their
priorities, the committee switched the Minnesota/MSU pairing with the
Miami/Cornell one to minimize the number of intraconference games at
one:
5W Mich State (C) 6E Denver U (W)
4W Minnesota (W) 3E Vermont (E)
1W Michigan (C) --+--2E Boston Univ (H)
|
2W North Dakota (W) --+--1E Clarkson (E)
3W Miami (C) 4E New Hampshire (H)
6W Cornell (E) 5E CO College (W)
1998:
Six Western teams (3 WCHA, 3 CCHA), Six Eastern teams (3 ECAC, 3 HE)
Let's look at this one for completeness. The reasons for this
year's decisions, according to the NCAA, are explained in more detail
in a separate post.
Esstern Qualifiers
Team lPWR RPI Comps Won
1 Boston Univ (H) 5 .595 BCCkNHYaPn
2 Boston Coll (H) 4 .581 CkNHYaPn
3 Clarkson (E) 3 .570 __ NHYaPn
4 New Hampshire (H) 2 .559 ____ YaPn
5 Yale (E) 1 .543 ______ Pn
6 Princeton (E) 0 .518 ________
The numbers say to send Yale and Princeton West, although one
would expect a desire to avoid conference matchups would result in two
teams from each league staying in their region. However, the
committee opted against this because it would have meant shipping out
UNH, a much bigger draw, in place of Yale. Also they were not that
concerned with the BC-UNH matchup in the second round, since the two
did not meet in the Hockey East playoffs, and split their season
series 1-1-1. (Yes, getting knocked out of the HE playoffs in the
first round actually helped UNH stay in the East, although I get the
feeling the attendance factor would have done the job anyway.) The
fact that UNH ends up being a 5-seed, and thus needs a nominal upset
to make the second round, might also have played a part. (The last
statement is pure speculation on my part, while the other two reasons
actually come from the NCAA.)
Western Qualifiers
Team lPWR RPI Comps Won
1 Mich State (C) 5 .621 NDMiOSWiCC
2 North Dakota (W) 4 .616 MiOSWiCC
3 Michigan (C) 3 .583 __ OSWiCC
4 Ohio State (C) 2 .565 ____ WiCC
5 Wisconsin (W) 1 .560 ______ CC
6 CO College (W) 0 .548 ________
Here, once again, going by the numbers means putting three
teams from the same conference in one region. This time the balance
is not tilted so far in favor of ignoring conference considerations:
Ohio State and Wisconsin are not that different as potential draws in
Ann Arbor, and the OSU-MSU second-round game would be a rematch of the
CCHA championship game, which the Spartans won after going 2-1 against
Ohio State in the regular season. So the committee did think about
shipping OSU in place of Wisconson, but opted against it. Why?
Perhaps the Eastern decision already set the tone of favoring
comparisons over conference considerations. At any rate, we learn
that avoiding intraconference matchups is not as essential as we
thought, and the following teams end up in each regional:
Team lPWR RPI Comps | Team lPWR RPI Comps
1 Mich State (C) 1 .621 ND | 1 Boston Univ (H) 1 .595 BC
2 North Dakota (W) 0 .616 | 2 Boston Coll (H) 0 .581
3 Michigan (C) 3 .583 OSYaPn | 3 Clarkson (E) 3 .570 WiNHCC
4 Ohio State (C) 2 .565 YaPn | 4 Wisconsin (W) 2 .560 NHCC
5 Yale (E) 1 .543 __ Pn | 5 New Hampshire (H) 1 .559 __ CC
6 Princeton (E) 0 .518 ____ | 6 CO College (W) 0 .548 ____
shuffling the seeds within the regionals would not change the
situation: these regionals will have one potential intraconference
game in each region.
5W Yale (E) 6E CO College (W)
4W Ohio State (C) 3E Clarkson (E)
1W Mich State (C) --+--2E Boston Coll (H)
|
2W North Dakota (W) --+--1E Boston Univ (H)
3W Michigan (C) 4E Wisconsin (W)
6W Princeton (E) 5E New Hampshire (H)
John Whelan, Cornell '91
<[log in to unmask]>
<http://www.cc.utah.edu/~jtw16960/joe.html>
Learn about the NCAA selection process on the web at
http://www.slack.net/~whelan/cgi-bin/tbrw.cgi?pairwise
HOCKEY-L is for discussion of college ice hockey; send information to
[log in to unmask], The College Hockey Information List.
|
