|
|
|
|
NCAA Tournament Bubble Watch - Statistical Approach
NOTE: THESE ARE THE PREDICTIONS FOR THE 2004 SEASON. BECAUSE THE RPI FORMULA WAS ALTERED,
THE MODEL COULD NOT BE RUN IN 2005.
As March arrives each year, much of the discussion among hardcore
NCAA basketball fans revolves around the proverbial NCAA Tournament "bubble." The big question is not
who will win their conference tournaments, but which 65 teams will gain a spot in the NCAA Tournament.
By the time the conference tournaments end on Selection Sunday, the suspense will be all but over for
about 55 to 60 teams. That includes teams that have gained automatic bids by winning their conference
tournament (or regular season championship in the case of the Ivy League) or are clear-cut locks based
on their won-loss record. That normally leaves about 15 to 20 teams squarely on the infamous bubble.
Those schools and their fans will be forced to squirm as the brackets are unveiled and we learn the
identity of the fortunate 65.
The process by which the NCAA Men's Basketball Tournament Selection Committee chooses the 34 at-large
bids (those teams that did not gain automatic entry by winning their conference championship) is not an
exact science, but the major selection criteria are as follows:
- RPI --
The RPI (Rating Percentage Index) is a statistical tool used by the Selection Committee to rank every
Division 1 school in America. The index takes into account each team's winning percentage, the winning
percentage of its opponents and the winning percentage of its opponents' opponents. The Committee relies
heavily on the RPI which rewards teams that play a difficult schedule and penalizes teams that line up
creampuff opponents outside of conference play. To cite an example, the Georgia Bulldogs were 16-14 two
years ago, but were invited to the tournament because they played the toughest schedule in America and
were ranked 27th in the RPI standings. Over the past nine years, no eligible team with an RPI below 33 has
been left out of the tournament. By contrast, only three teams with an RPI of 60 or above have made the
tournament during that time period. Those teams were Wyoming (63rd in 2002), Minnesota (66th in 1995) and
New Mexico (74th in 1999). New Mexico's invitation was almost certainly made possible by the fact that
their Athletic Director and League Commissioner were on the Selection Committee. It is highly
unlikely that a team with an RPI of 70 or higher will make the tournament this season. The following
table shows just how important RPI has been over the past nine years:
RPI Rank | % of Eligible Teams Invited to Tournament 1995-2003 |
0 to 29 | 100% |
30 to 39 | 93% |
40 to 49 | 56% |
50 to 59 | 25% |
60 to 69 | 3% |
* Excluding Automatic Bids |
- Conference Winning Percentage/Standings --
In the six major conferences (ACC, Big East, Big Ten, Big XII, SEC, Pac Ten) and even the second-tier
conferences (Atlantic 10, Conference USA, Mountain West, WAC), a team's record in conference play is
one the most important indicators as to whether or not the team will be invited to the "Big Dance."
Rarely will the NCAA Selection Committee invite a team that finishes below .500 in regular season
conference play. It is also rare for the Committee to bypass a team in the league standings (or division
within a league) to invite another team that finished lower in the standings. Since 1996, the only time
that a team in one of the six major conferences was bypassed was in 1998 when Florida State (6-10 in ACC
play) was invited and 7-9 Wake Forest was not. This was also the only time since 1996 that a team who
finished four games below .500 in conference play received an at-large bid to the NCAA Tournament.
Since 1995, 92% of teams that finished above .500 in one of the six top conferences (excluding automatic
qualifiers) were invited to the Tournament (193 of 210). Of the 17 that did not, only six had an RPI
ranking better than 60 (four of those were Big East teams). By contrast, only 30% of the teams
in those six conferences that finished exactly .500 in league play earned a bid. This data is a bit
deceiving, however, because only 1 of 18 Big East teams that finished exactly .500 received a spot in
the NCAA Tournament. In the other five major conferences, 14 of 32 teams with a .500 conference record
received a spot (including 4 of 5 ACC teams). Moreover, only six times in eight years has a team finished below .500
and been invited (three ACC teams, two Big Ten teams and last year's shocker, Alabama). Needless to say,
there seems to be a severe conference bias where the NCAA Tournament is concerned. A final note on
conference winning percentage: between 1995 and 2003, only four teams from one of the six major conferences
finished four games above .500 in conference play and were not invited to the Tournament (West Virginia,
11-7 in 1997; Nebraska, 10-6 in 1999; Boston College and Seton Hall, 10-6 in 2003).
- Road/Neutral Court Record --
Because no one plays home games in the NCAA Tournament, the Selection Committee looks closely at a
team's road and neutral court record as an indicator of how well they will perform in the NCAA Tournament.
The Committee will look favorably on teams that have shown the ability to win on the road and may look
unfavorably on a team with a poor record away from home.
- Record in the Last 10 Games --
The Selection Committee loves a team that rolls into March Madness on a hot streak. A team in a major
conference with 23 wins will certainly not be excluded from the tournament if they finish the season
3-7, but for a team on the bubble, late season performance could mean the difference
between a spot in the field of 65 and a trip to the NIT.
- Record Against the RPI Top 50 --
Typically, more than 90% of the RPI Top 50 earns a bid to the NCAA Tournament so a team's record against
this group is a good indicator of how they will stack up against the field of 65. The Committee pays
very close attention to road wins against teams with strong RPI rankings.
- Total Wins --
It was once believed that a 20-win season in a major conference would guarantee a spot in the NCAA
Tourney. This is no longer the case as Syracuse found out in 2002 and Alabama learned in 2001.
Twenty wins in a major conference makes a strong statement but it will not, by itself, earn
you a dance card. At one time, teams from small conferences could only gain entry to the NCAA
Tournament by winning their conference championship. Now, these so-called mid major programs are
competing for at-large bids with middle of the pack teams from the major conferences. Last year,
Southern Illinois, Butler and Gonzaga each grabbed at large bids. In 2002, it was Southern Illinois
and Pepperdine.
Unfortunately, many of the talking heads that we see on television this time of year making
predictions about who will and will not make the tournament don't seem to have a firm grasp on the
Committee's criteria. If you listen to Dick Vitale, you'd swear that there are 90 teams in the Tournament.
If you listen to Jay Bilas, you might think that there are only 50 teams invited and that the Big East
is only allowed two. The major sports websites like CBSSportsline and ESPN don't seem to be very
accurate either. One very good source for bubble information is Jerry Palm's collegerpi.com. The man
knows his March Madness and is more accurate at predicting the field than anyone else I've seen.
So you might ask, who will be the 65 teams invited to this year's NCAA Tournament or more specifically,
who will be the 34 at-large teams joining the 31 conference champions? To answer that question, I have
taken a new approach - a purely statistical one. Using data from 1995 to 2003, I have created econometric
models (logistic models for those of you who care) that are designed to help predict the 34 at-large
teams. The data that I have collected includes the following: RPI Ranking, Total Wins, Overall Winning
Percentage, Conference Winning Percentage, Record in the Last Ten Games, Road/Neutral Court Record, Record
Against the RPI Top 50 and Conference RPI (for the six top conferences). This year I added a variable to
capture conference bias (more on that below). I then segmented the teams into three groups: those in the
six major conferences (ACC, Big East, Big Ten, Big XII, SEC, Pac Ten), those in the second-tier conferences
(Atlantic 10, Conference USA, Mountain West, WAC) and the rest. This segmentation was necessary
because teams from the major conferences are evaluated much differently than teams in small conferences.
For example, a .500 record in the SEC clearly holds more weight than a .500 record in the WAC. I used
these models to predict the odds that each team would be invited to the NCAA Tournament based on past
data. I evaluated all teams with RPI rankings of less than 80, excluding those teams currently leading
their conference. According to the analysis, the following variables proved most statistically significant
and thus were included in the final prediction models:
- Big Six Conferences -- RPI Ranking, Overall Winning Pct, Conference Winning Pct, Winning Pct Against the RPI Top 50, ACC variable*, Big East variable*
- Second-Tier Conferences -- RPI Ranking, Overall Wins, Conference Winning Percentage
- Small Conferences -- RPI Ranking, Conference Winning Percentage
This is not to say that wins in the last ten games and road/neutral court record are not important in
the eyes of the Selection Committee, but that the variables above produced the best statistical models.
* This year, I added a variable that tested for conference bias. As I mentioned above, the Committee
appears to favor the ACC and Big Ten, often at the expense of the Big East. Last year, both Seton Hall
and Boston College were snubbed. Seton Hall finished 10-6 in Big East play, ranked 42nd in RPI and
won 10 of their final 13 games. Boston College won its division with a 10-6 record, was a respectable
49th in RPI ranking and also won 10 of 13 down the stretch. Meanwhile, Alabama finished 7-9 in the SEC,
lost 9 of 15 down the stretch and didn't even make it to the quarterfinals of the SEC Tournament.
Their RPI ranking was 38th, hardly strong enough given how poorly they graded in all other important
criteria. The only explanation for Alabama's inclusion in the NCAA Tournament is that someone from Bama
has incriminating pictures of someone on the NCAA Tournament Selection Committee. The Boston College
players and coaches also had to be shaking their heads when NC State received a bid. The Eagles finished
with a better RPI than the Wolfpack, a better conference record, a better record in the last ten games,
a better road and neutral record and even beat NC State on its own home court. NC State received the
bid only because they play in the more popular ACC. Thankfully for BC fans, the Eagles will be joining
the ACC in 2005-06 and may someday be able to take advantage of the bias, rather than be penalized by it
as they were last season.
To test for this suspected bias, I added a variable for each of the six major conferences.
My analysis found that there was in fact a bias in favor of the ACC and a negative bias against the
Big East. The results for these two conferences were statistically significant (using a confidence
level of 90%). I was not able to prove bias for or against the other four conferences based on the
data. Thus, only the ACC and Big East variables were included in the final model. I also included
each conference's RPI and RPI ranking in the initial models so that I could test whether or not it was
the conference's RPI and not simply their name recognition that affected whether or not they would
receive an invitation to the NCAA Tournament. The Conference RPI variables did not prove to be
statistically significant.
So here are the current probabilites that each team will be in the NCAA Tournament according to my
model and data through Saturday, March 13th.
Last year, my model correctly predicted 32 of 34 at-large teams. The 2003 model predicted that
Seton Hall and Boston College would make the tournament and that Indiana and Alabama would not.
Despite the results of my model, I fully expected Indiana to be in the field of 65. You could have
knocked me over with a feather when I saw the name "Alabama" appear on the brackets.
RPI |
School |
Tot Wins |
Win % |
Conf Win % |
Road/N Win % |
Wins L10 |
Win %
RPI-T50 |
PROB of ALB |
1 |
Duke |
27 |
87.1% |
83.3% |
81.3% |
7 |
76.9% |
100.0% |
4 |
MississippiSt |
25 |
89.3% |
82.4% |
92.9% |
8 |
75.0% |
100.0% |
7 |
Pittsburgh |
28 |
87.5% |
78.9% |
76.9% |
7 |
69.2% |
100.0% |
8 |
Texas |
23 |
79.3% |
77.8% |
61.5% |
7 |
69.2% |
100.0% |
3 |
SaintJosephs |
27 |
96.4% |
94.1% |
94.1% |
9 |
83.3% |
100.0% |
21 |
Illinois |
24 |
82.8% |
83.3% |
75.0% |
10 |
50.0% |
100.0% |
17 |
NorthCarolinaSt |
20 |
69.0% |
66.7% |
46.2% |
6 |
50.0% |
100.0% |
13 |
Kansas |
20 |
71.4% |
72.2% |
53.3% |
6 |
33.3% |
100.0% |
14 |
GeorgiaTech |
23 |
71.9% |
55.6% |
62.5% |
6 |
53.3% |
100.0% |
12 |
Florida |
20 |
69.0% |
61.1% |
64.3% |
7 |
50.0% |
100.0% |
15 |
Syracuse |
21 |
75.0% |
64.7% |
63.6% |
7 |
46.2% |
100.0% |
20 |
Providence |
20 |
71.4% |
64.7% |
69.2% |
6 |
66.7% |
100.0% |
19 |
WakeForest |
19 |
67.9% |
52.9% |
53.8% |
6 |
53.3% |
100.0% |
22 |
BostonCollege |
23 |
71.9% |
63.2% |
62.5% |
8 |
50.0% |
100.0% |
18 |
NorthCarolina |
18 |
64.3% |
47.1% |
42.9% |
5 |
50.0% |
100.0% |
27 |
Vanderbilt |
21 |
70.0% |
52.6% |
46.2% |
6 |
46.2% |
99.9% |
40 |
Dayton |
24 |
75.0% |
73.7% |
64.3% |
5 |
50.0% |
99.9% |
30 |
TexasTech |
22 |
68.8% |
55.6% |
40.0% |
5 |
36.4% |
99.8% |
32 |
Memphis |
21 |
75.0% |
70.6% |
46.2% |
7 |
37.5% |
99.7% |
28 |
SetonHall |
19 |
67.9% |
58.8% |
46.2% |
6 |
40.0% |
99.7% |
34 |
Charlotte |
21 |
72.4% |
72.2% |
68.8% |
7 |
50.0% |
99.7% |
25 |
SouthernIll |
25 |
86.2% |
90.0% |
80.0% |
8 |
0.0% |
99.6% |
37 |
DePaul |
21 |
70.0% |
73.7% |
66.7% |
8 |
45.5% |
99.6% |
33 |
UAB |
20 |
69.0% |
72.2% |
42.9% |
7 |
36.4% |
99.5% |
39 |
MichiganSt |
18 |
62.1% |
72.2% |
40.0% |
7 |
10.0% |
99.1% |
45 |
SouthCarolina |
23 |
69.7% |
52.6% |
56.3% |
4 |
60.0% |
98.9% |
31 |
BYU |
19 |
70.4% |
68.8% |
50.0% |
9 |
33.3% |
98.8% |
41 |
Arizona |
20 |
69.0% |
60.0% |
50.0% |
6 |
25.0% |
98.5% |
24 |
Louisville |
20 |
69.0% |
55.6% |
50.0% |
4 |
54.5% |
98.1% |
26 |
Alabama |
17 |
58.6% |
50.0% |
42.9% |
6 |
26.7% |
94.4% |
38 |
LSU |
18 |
64.3% |
47.1% |
46.2% |
4 |
50.0% |
93.9% |
43 |
UtahSt |
24 |
88.9% |
89.5% |
76.9% |
8 |
50.0% |
90.9% |
62 |
Washington |
19 |
63.3% |
66.7% |
50.0% |
8 |
50.0% |
89.4% |
70 |
AirForce |
22 |
78.6% |
80.0% |
60.0% |
7 |
75.0% |
76.7% |
49 |
Richmond |
20 |
62.5% |
63.2% |
57.9% |
7 |
16.7% |
68.8% |
46 |
UTEP |
22 |
75.9% |
71.4% |
57.1% |
7 |
25.0% |
41.1% |
56 |
Colorado |
17 |
63.0% |
58.8% |
46.2% |
6 |
22.2% |
34.6% |
48 |
Oklahoma |
19 |
65.5% |
50.0% |
46.2% |
5 |
20.0% |
34.1% |
64 |
Rice |
21 |
67.7% |
65.0% |
58.8% |
7 |
28.6% |
27.1% |
50 |
NotreDame |
17 |
58.6% |
55.6% |
53.8% |
7 |
33.3% |
12.5% |
66 |
TroySt |
23 |
82.1% |
87.0% |
77.8% |
9 |
0.0% |
12.3% |
51 |
Missouri |
16 |
55.2% |
55.6% |
40.0% |
7 |
18.2% |
8.9% |
55 |
Michigan |
18 |
62.1% |
50.0% |
38.5% |
5 |
25.0% |
8.1% |
47 |
Georgia |
16 |
55.2% |
44.4% |
23.1% |
5 |
33.3% |
5.1% |
52 |
Virginia |
17 |
58.6% |
38.9% |
33.3% |
5 |
21.4% |
4.0% |
53 |
FloridaSt |
17 |
56.7% |
35.3% |
21.4% |
3 |
26.7% |
1.5% |
65 |
SaintLouis |
18 |
60.0% |
57.9% |
40.0% |
7 |
30.8% |
1.4% |
74 |
GeorgeWash |
18 |
62.1% |
66.7% |
40.0% |
7 |
27.3% |
1.3% |
75 |
WisMilwaukee |
18 |
64.3% |
77.8% |
56.3% |
5 |
0.0% |
1.0% |
69 |
WichitaSt |
20 |
66.7% |
65.0% |
62.5% |
7 |
25.0% |
0.6% |
76 |
KentSt |
21 |
72.4% |
71.4% |
64.7% |
5 |
33.3% |
0.4% |
72 |
BoiseSt |
19 |
67.9% |
65.0% |
66.7% |
8 |
50.0% |
0.4% |
73 |
Hawaii |
18 |
62.1% |
57.9% |
38.5% |
3 |
33.3% |
0.3% |
68 |
IowaState |
16 |
57.1% |
44.4% |
15.4% |
4 |
37.5% |
0.2% |
78 |
Creighton |
20 |
71.4% |
63.2% |
46.2% |
4 |
0.0% |
0.1% |
57 |
Rutgers |
15 |
55.6% |
41.2% |
23.1% |
4 |
20.0% |
0.0% |
80 |
UNLV |
17 |
58.6% |
52.9% |
42.9% |
6 |
0.0% |
0.0% |
77 |
Tennessee |
15 |
53.6% |
41.2% |
9.1% |
4 |
25.0% |
0.0% |
67 |
Villanova |
15 |
50.0% |
42.1% |
52.9% |
3 |
21.4% |
0.0% |
79 |
Auburn |
14 |
50.0% |
29.4% |
18.2% |
3 |
25.0% |
0.0% |
|
|
|
|
|
|
|
|
|
Teams in the shaded area will be invited to the NCAA Tournament, according to the model.
Prob of ALB means probability of an at-large bid according to the model.
Teams in the RPI Top 80 currently leading their conferences weren't included in the analysis
|
|
|
ESPN College Hoops
CBSSportsline College Hoops
CollegeRPI.com
StatFox.com
StatSheet.com
|
|