Tuesday, March 19, 2019

How Well Did The Computers Predict The Field?

All numbers below are as of Monday morning (i.e. they include all of the results up through Selection Sunday but do not include any post-Selection Sunday tournaments).


NET

Ten highest rated teams to miss the Tournament (NIT seed given):

33. NC State (2)
35. Clemson (2)
38. Texas (2)
41. Furman (3)
46. Memphis (3)
48. Nebraska (4)
49. Lipscomb (5)
50. Penn St (-)
52. TCU (1)
53. Creighton (2)

Ten lowest rated teams to earn an at-large (seed given):
73. St. John's (11)
63. Arizona St (11)
61. Minnesota (10)
57. Seton Hall (10)
56. Temple (11)
55. Ohio St (11)
47. Belmont (11)
45. Washington (9)
43. Iowa (10)
42. Syracuse (8)

  BPI Strength of Record

Ten highest rated teams to miss the Tournament (NIT seed given):

31. NC State (2)
33. UNC-Greensboro (1)
43. Clemson (2)
44. TCU (1)
46. Indiana (1)
51. Nebraska (4)
55. Alabama (1)
56. Furman (3)
57. Texas (2)
58. Creighton (2)

Ten lowest rated teams to earn an at-large (seed given):
54. Arizona State (11)
53. St. John's (11)
50. Florida (10)
48. Baylor (9)
47. Temple (11)
45. Mississippi (8)
41. Seton Hall (10)
40. Belmont (11)
39. VCU (8)
38. UCF (9)


Pomeroy

Ten highest rated teams to miss the Tournament (NIT seed given):

29. Clemson (2)
30. Texas (2)
33. NC State (2)
39. Nebraska (4)
40. Penn St (-)
42. Indiana (1)
48. TCU (1)
50. Creighton (2)
53. Lipscomb (5)
56. Furman (3)

Ten lowest rated teams to earn an at-large (seed given):
78. St. John's (11)
76. Temple (11)
61. Arizona St (11)
55. Seton Hall (10)
54. Belmont (11)
51. Washington (9)
47. Minnesota (10)
46. UCF (9)
45. Ohio St (11)
44. Mississippi (8)

--------------------------------

Did the NET change anything?
Not really. Remember that in the old days, nobody ever selected or seeded teams by the RPI itself. Rather they used the RPI SOS and record vs RPI Top 50/100/etc. So yet again, teams were not really chosen by their NET (the ESPN Strength of Record was a better predictor), but teams got in because of record vs NET Quadrant 1/2 instead of RPI Top 50/100. For example, St. John's had nothing going for their resume other than five Quadrant 1 wins. Seton Hall, Arizona State, and Minnesota also stand out as getting in over Quadrant 1 and/or Quadrant 2 records.

The one improvement of the NET is that it's not as easily manipulable, and it doesn't so badly punish teams for some garbage teams on their schedule. The fact that NC State was even in the at-large consideration was an improvement, as the old RPI would have dumped them all the way to 97th, far outside where they would have been considered. The NET acknowledge NC State's garbage non-conference schedule, ranking it dead last in the nation (353rd), but still rated them 33rd, keeping them in the at-large discussion until the final day. NC State's athletic director whined about them being left out despite a high NET, but that's just a misunderstanding of how computer ratings have ever been used.

How was the overall bracket?
Honestly, good. There are some seedings worth critiquing, of course. Florida State clearly deserved a 3 seed but got dumped to a 4 in a rough pod. Wisconsin clearly deserved a 4 seed and not only got a 5 seed but got one of the hottest teams in the country in Oregon. Michigan State clearly deserved a 1 seed and not only got a 2 but drew Duke.

But all of these are teams just one seed line off, and it's hard to complain too much about that. The days of teams getting mystifying seeds three or four lines off of where they deserved seem to be over.

Why? Because everybody reads social media now. The Selection Committee members can all read the Bracket Matrix and know that they're going to get creamed on social media if they whiff badly. I was arguing on Sunday that Belmont would get an at-large bid specifically because of social media - that they had become the avatar for those frustrated by the lack of mid-major representation in the at-large bids.

The fact that a team like UNC-Greensboro got left out and that a team like Furman was not seriously considered means that anti-mid major bias still exists, but it definitely has gotten way, way better, just in the past three years. That's something to celebrate.