Tuesday, March 15, 2016

Which Team Got Screwed Over the Most?

The NCAA tournament selection committee has had its say, and now the predicting has come to an end, not just for this site but for everyone. This year I found 6 bracket projections (which take into account bracketing principles) and 25 ranked lists for a total of 31 college women's basketball predictions. Most ranked lists exist for purposes other than bracket prediction (power rankings, score prediction ratings), but they can be used predictively to develop a bracket just the same, and so I include them.

The bona fide bracket predictions were developed by the following people:

  1. Charlie Creme - ESPN.com
  2. Matt Olson - Omnirankings.com
  3. CSM Staff - CollegeSportsMadness.com
  4. Matt5762 - boards.rebkell.net
  5. Whoever is in charge of RealTimeRPI.com's bracket prediction
  6. Steven Bell - this site

These are the 25 ranked lists I included in this analysis.

  1. S-Factor (current version)
  2. Old S-Factor (pre-2015 version)
  3. Charlie Creme without procedural bumps -ESPN.com
  4. D1SN "poll" - William Lansdale -D1sportsnet.com
  5. AP poll
  6. Coaches' poll
  7. SpaceJunkie - boards.rebkell.net
  8. RealTimeRPI RPI - realtimerpi.com
  9. RealTimeRPI power - realtimerpi.com
  10. Sagarin Rating - http://www.rpiratings.com/womrate.php
  11. Sagarin GoldenMean
  12. Sagarin Predictor
  13. Sagarin Recent
  14. Massey Rating - masseyratings.com
  15. Massey Power Rating
  16. Omni Rankings - omnirankings.com
  17. Warren Nolan Elo Chess
  18. Warren Nolan's NPI - http://warrennolan.com/basketballw/2015/npi
  19. Sonny Moore's Computer Power Ratings - http://www.sonnymoorepowerratings.com
  20. Hero Sports' Bennett Rank - http://herosports.com/colleges/sports/d1-womens-basketball
  21. Georgia Tech - LRMC rankings - Bayesian - http://www2.isye.gatech.edu/~jsokol/lrmcW/
  22. Georgia Tech - LRMC rankings - Classic
  23. Georgia Tech - LRMC rankings - LRMC(0)
  24. NCAA Win-Loss percentage
  25. NCAA RPI
On each of these ranked lists, I found the 64 teams that would have been predicted by said list - 32 at-large bids and 32 automatic bids. Some of these lists produce rankings similar to how the selection committee examines teams, and some produce wild and unrealistic lists. These outlier lists (Warren Nolan Elo Chess, NPI, W-L% and the polls, which don't produce full lists) were not included when I found the average predicted seed for each team.  

In the list below, the larger the magnitude of the number means the larger difference between the team's seed as predicted by these 31 ranked lists and brackets and the team's actual seed in the tournament this year. Negative numbers indicate teams seeded below their predicted seeding, and positive numbers indicate teams seeded above their predicted seeding.


Deviation from average
Average predicted seed
Actual seed
Team
-2.63
8.38
(11)
Princeton
-2.47
8.53
(11)
Purdue
-2.31
10.69
(13)
Army
-2.16
8.84
(11)
Colorado St.
-2.07
8.93
(11)
James Madison
-1.82
14.18
(16)
Idaho
-1.80
7.20
(9)
Duquesne
-1.72
8.28
(10)
Green Bay
-1.69
10.31
(12)
Albany
-1.50
13.50
(15)
New Mexico St.
-1.19
14.81
(16)
Jacksonville
-0.96
9.04
(10)
Missouri
-0.86
11.14
(12)
So. Dakota St.
-0.58
5.42
(6)
DePaul
-0.57
13.43
(14)
Cent. Arkansas
-0.48
3.52
(4)
Stanford
-0.46
14.54
(15)
Iona
-0.38
6.62
(7)
Washington
-0.29
1.71
(2)
Maryland
-0.28
6.72
(7)
Tennessee
-0.26
4.74
(5)
Florida St.
-0.18
9.82
(10)
St. Bonaventure
-0.14
9.86
(10)
Pennsylvania
-0.07
11.93
(12)
Mid. Tenn.
-0.03
1.97
(2)
Oregon St.
0.00
1.00
(1)
Baylor
0.00
1.00
(1)
Notre Dame
0.00
7.00
(7)
BYU
0.00
12.00
(12)
Chattanooga
0.00
14.00
(14)
Hawaii
0.00
16.00
(16)
Robert Morris
0.00
16.00
(16)
NC A&T
0.03
3.03
(3)
Ohio St.
0.03
6.03
(6)
West Virginia
0.06
1.06
(1)
Connecticut
0.07
15.07
(15)
Troy
0.07
6.07
(6)
Oklahoma
0.08
8.08
(8)
Georgia
0.13
5.13
(5)
Mississippi St.
0.14
13.14
(13)
Missouri St.
0.18
13.18
(13)
San Francisco
0.29
4.29
(4)
Syracuse
0.29
1.29
(1)
South Carolina
0.30
8.30
(8)
Geo. Wash.
0.32
2.32
(2)
Texas
0.42
3.42
(3)
Louisville
0.48
9.48
(9)
Indiana
0.50
6.50
(6)
South Florida
0.57
14.57
(14)
Buffalo
0.61
13.61
(13)
Belmont
0.65
3.65
(3)
UCLA
0.80
9.80
(9)
Kansas St.
0.89
14.89
(14)
UNC Asheville
0.96
15.96
(15)
Alabama St.
0.97
4.97
(4)
Michigan St.
1.04
8.04
(7)
Oklahoma St.
1.13
10.13
(9)
Auburn
1.13
4.13
(3)
Kentucky
1.55
3.55
(2)
Arizona St.
1.63
9.63
(8)
St. John's
1.71
6.71
(5)
Miami (FL)
1.83
5.83
(4)
Texas A&M
1.84
9.84
(8)
Seton Hall
2.43
7.43
(5)
Florida


For the second year in a row, the Most Screwed Team is Princeton. However, the magnitude is much smaller this year, and there are many (Charlie Creme included) who believe that Princeton is extremely lucky to receive a bid (the Ivy League's first at-large ever) at all. Second place Purdue was also on the bubble. Third place Army, which averaged a seed of 10.69, was given a 13 seed and might be considered Pretty Screwed Indeed. Florida, whose 5-seed was 2.43 seeds better than its average, is considered by this method to be the luckiest team. Overall though, the selection committee really conformed to expectations for the most part. 

However, this isn't the only method by which one can determine the Most Screwed Team. By finding the list of 32 most commonly chosen at-large teams in the tournament and comparing it to the list of the 32 actually selected at-large teams, we can find the People's Choice Award for the Most Screwed Team. This year there were three teams that would have been selected by a plurality of ranked lists that did not get into the tournament: Rutgers, Duke and Oregon.


Rank
no. of pred.
% of pred.
Actual seed
Team
1
31
100%
(2)
Texas
1
31
100%
(2)
Arizona St.
1
31
100%
(3)
UCLA
1
31
100%
(3)
Louisville
1
31
100%
(3)
Kentucky
1
31
100%
(3)
Ohio St.
1
31
100%
(4)
Michigan St.
1
31
100%
(4)
Stanford
1
31
100%
(4)
Syracuse
1
31
100%
(5)
Mississippi St.
1
31
100%
(5)
Florida St.
1
31
100%
(5)
Miami (FL)
1
31
100%
(6)
DePaul
14
30
97%
(5)
Florida
14
30
97%
(6)
South Florida
14
30
97%
(6)
West Virginia
17
29
94%
(4)
Texas A&M
18
27
87%
(6)
Oklahoma
19
26
84%
(7)
Washington
19
26
84%
(8)
Georgia
21
25
81%
(7)
Oklahoma St.
21
25
81%
(7)
Tennessee
21
25
81%
(10)
Missouri
24
24
77%
(11)
Princeton
25
23
74%
(9)
Indiana
26
20
65%
(7)
BYU
26
20
65%
(9)
Duquesne
28
19
61%
(8)
Seton Hall
28
19
61%
(11)
Purdue
30
14
45%
Rutgers
31
13
42%
Duke
32
12
39%

Oregon
33
11
35%
(10)
St. Bonaventure
34
9
29%
USC
35
8
26%
(9)
Auburn
35
8
26%
Fla. Gulf Coast
35
8
26%
Arkansas St.
38
7
23%
N.C. State
38
7
23%
UTEP
38
7
23%
W. Kentucky
41
6
19%
Saint Louis
41
6
19%
Michigan
41
6
19%
St. Mary's
44
5
16%
(9)
Kansas St.
44
5
16%
Iowa
44
5
16%
Temple
47
4
13%
Maine
47
4
13%
Ohio
47
4
13%
Northwestern
50
3
10%
South Dakota
50
3
10%
UC Riverside
52
2
6%
VCU
52
2
6%
Quinnipiac
52
2
6%
San Diego
52
2
6%
Georgia Tech
56
1
3%
Tulane
56
1
3%
Bucknell
56
1
3%
Long Beach St.


The teams left off the People's Choice list that actually did make it were St. Bonaventure, Auburn, and (selected by only 5 predictions) Kansas State. By this method, one could say that Kansas State's selection was as likely as if Iowa or Temple got in, and slightly less likely than St. Mary's, Michigan or Saint Louis.