Monday, December 2, 2019

THE AT LARGE SELECTIONS: WHO GETS THE LAST TWO SLOTS?

After the seeding, the Committee had 23 at large positions left to fill.  Here is how my computer had the remaining teams from the Top 60 lined up for at large selections:

#15 Virginia Tech met 39 "yes" at large standards and 0 "no
#14 Duke  37 and 0
#30 California  28 and 0
#33 Vanderbilt  26 and 0
#21 Texas A&M  22 and 0
#28 Louisville  22 and 0
#19 NC State  18 and 0
#23 Florida  14 and 0
#27 Washington State  14 and 0
#36 Clemson  14 and 0
#25 Michigan  11 and 0
#29 Santa Clara  11 and 0
#46 Pepperdine  11 and 0
#26 Memphis  10 and 0
#31 West Virginia  9 and 0
#35 Georgetown  9 and 0
#32 Florida Atlantic  8 and 0
#34 Notre Dame  8 and 0
#40 Arizona  8 and 0
#53 Tennessee  4 and 3
#43 Colorado  1 and 0
#48 Texas  0 and 0
#55 TCU  0 an 0
****************************
#39 Alabama  0 and 0
#45 Iowa  0 and 0
#57 Georgia  2 and 3
#41 Harvard  1 and 2
#49 Columbia  0 and 1
#51 Utah  0 and 1

My system picked the top 23 teams on the list as the at large selections, on the basis of their "yes" and "no" results, with Tennessee as a pick because of its net +1 "yes" result.  For the 0 and 0 teams, my system picked Texas and TCU over Alabama and Iowa because Texas and TCU had better Top 50 Results scores.

My system's picks matched the Committee's for all but 2 teams.  The yellow highlighted teams are the ones where the Committee's teams differed from my system's.

In my first article on Brown, Yale, Harvard, Columbia, and the Ivy League as a whole, I discussed Harvard and Columbia, so I won't write much here about the likely reasons they didn't get at large spots.

Here are some details about Florida Atlantic, Tennessee, Alabama, Iowa, Georgia, and Utah.

#32 Florida Atlantic

Florida Atlantic met 8 "yes" at large selection standards and 0 "no" standards.  Here are its "yes" standards:

RPI (Standard #1)

The "yes" at large selection score for this standard is >=.5989.

Florida Atlantic's RPI was .6045.  When I change the "yes" score for this standard to accommodate Florida Atlantic's not getting an at large selection, there will be 34 teams, of 404 unseeded at large selections over the last 12 years, that previously were in the "yes" at large selection category but that now will move out of it.

RPI and RPI Rank (Standard #14)

The "yes" at large selection score for this standard is >=3.3846.

Florida Atlantic's RPI Rank was #32, combining with its RPI for a score of 3.4162.  When I change the "yes" score for this standard, there will be 33 teams, of 404, that previously were in the "yes" at large selection category that now will move out of it.

RPI and Non-Conference RPI Rank (Standard #16)

The "yes" at large selection score for this standard is >=3.4543.

Florida Atlantic's Non-Conference RPI Rank was #34, combining with its RPI for a score of 3.4748.  When I change the "yes" score for this standard, there will be 17 teams, of 404, that previously were in the "yes" at large selection category that now will move out of it.

RPI and Top 50 Results Rank (Standard #18)

The "yes" at large selection score for this standard is >=3.704.

Florida Atlantic's Top 50 Results Rank was #56, combining with its RPI for a score of 3.4028.  When I change the "yes" score for this standard, there will be 35 teams, of 404, that previously were in the "yes" at large selection category that now will move out of it.

RPI and Common Opponents Results Rank (Standard #24)

The "yes" at large selection score for this standard is >=3.3867.

Florida Atlantic's Common Opponents Rank was #38, combining with its RPI for a score of 3.4113.  When I change the "yes" score for this standard, there will be 22 teams, of 404, that previously were in the "yes" at large selection category that now will move out of it.

RPI Rank and Conference Standing (Standard #30)

The "yes" at large selection score for this standard is <=43.8.

Florida Atlantic's RPI Rank was #32 and its Conference Standing was 1.5, combining for a score of 40.9.  When I change the "yes" score for this standard, there will be 19 teams, of 404, that previously were in the "yes" at large selection category that now will move out of it.

Non-Conference RPI and Conference Standing (Standard #40)

The "yes" at large selection score for this standard is >=2.4125.

Florida Atlantic's Non-Conference RPI was .6123, which combined with its Conference Standing for a score of 2.4424.  When I change the "yes" score for this standard, there will be 10 teams, of 404, that previously were in the "yes" at large selection category that now will move out of it.

Non-Conference RPI Rank and Conference Standing (Standard #49)

The "yes" at large selection score for this standard is <=77.5.

Florida Atlantic's score for this standard was 67.0.  When I change the "yes" score for this standard, there will be 31 teams, of 404, that previously were in the "yes" at large selection category that now will move out of it.

Comments on Florida Atlantic

The changes from the Committee's pattern, in not giving Florida Atlantic an at large selection, are not great.  There's still a question, however, of "Why?"

For an answer to that question, I first look at Florida Atlantic's Top 50 Results.  They were a loss to #23 Florida at home; a loss to #9 Kansas away; a win against #47 North Texas (Conference USA) at home; and a loss to #47 North Texas away.  It's one good Top 50 result gave it, in my system, a Top 50 Results score of 2 and Rank of #56.  Of all the teams in the list at the beginning of this article -- likely all of the teams the Committee was considering -- this was the poorest results against Top 50 teams.

All of Florida Atlantic's "yes" standards involved either its RPI or its Non-Conference RPI.  It seems likely that the Committee, after looking at Florida Atlantic's poor Top 50 Results, decided that it didn't believe its RPI and Non-Conference RPI.  Thus it appears Florida Atlantic's poor Top 50 results kept it from getting an at large selection.  This seems like a reasonable decision.

#53 Tennessee

Tennessee met 4 "yes" at large selection standards and 3 "no" standards, meaning it had a profile the Committee hasn't seen over the last 12 years.

Here are Tennessee's "yes" standards:

Top 50 Results Rank (Standard #6)

The "yes" at large selection score for this standard is <=12.

Tennessee's Top 50 Results Rank was 10.  When I change the "yes" score for this standard, there will be 30 teams, of 404 unseeded at large selections over the last 12 years, that previously were in the "yes" at large selection category that now will move out of it.

Non-Conference RPI Rank and Top 50 Results Rank (Standard #48)

The "yes" at large selection score for this standard is <=83.2.

Tennessee's Non-Conference RPI Rank was #31, which combined with its Top 50 Results Rank to give it a score for this standard of 67.0.  When I change the "yes" score for this standard, there will be 30 teams, of 404, that previously were in the "yes" at large selection category that now will move out of it.

Top 50 Results Score and Top 50 Results Rank (Standard #56)

The "yes" at large selection score for this standard is >= 21,275.

Tennessee's Top 50 Results Score was 14,493, which combine with its Top 50 Results Rank to give it a score for this standard of 21,493.  When I change the "yes" score for this standard, there will be 3 teams, of 404, that previously were in the "yes" at large selection category that now will move out of it.

Top 50 Results Rank and Conference Rank (Standard #66)

The "yes" at large selection score for this standard is <=17.8.

Tennessee's SEC Conference Rank was #3, which combined with its Top 50 Results Rank to give it a score for this standard of 16.3.  When I change the "yes" score for this standard, there will be 6 teams, of 404, that previously were in the "yes" at large selection category that now will move out of it.

Here are Tennessee's "no" standards:

RPI and Conference Standing (Standard #19)

The "no" at large selection score for this standard is <=3.0518.

Tennessee's RPI was .5734 and its Conference Standing was #11, combining for a score for this standard of 3.0154.  If Tennessee had gotten an at large selection, the required change in the "no" score for this standard would have caused 6 teams, of the 133 Top 60 teams not getting at large selections over the last 12 years, that previously were in the "no" at large selection category, to move out of that category.

RPI Rank and Conference Standing (Standard #30)

The "no" at large selection score for this standard is >=108.9.

Tennessee's RPI Rank was #53, which combined with its Conference Standing to give it a score for this standard of 117.9.  If Tennessee had gotten an at large selection the required change for this standard would have caused 7 teams, of 133, that previously were in the "no" at large selection category, to move out of it.

Non-Conference RPI Rank and Conference Standing (Standard #49)

The "no" at large selection score for this standard is >=272.0.

Tennessee's Non-Conference RPI Rank was #31, which combined with its Conference Standing to give it a score for this standard of 273.  If Tennessee had gotten an at large selection the required change for this standard would have caused 1 team, of 133, that previously was in the "no" at large selection category, to move out of it.

Comments on Tennessee

All of the "yes" at large selection standards Tennessee met included its Top 50 Results Score or Rank.  All of the "no" standards included its Conference Standing.  It seems clear the Committee felt Tennessee's Top 50 Results weren't good enough to overcome its #11 position in the SEC Conference Standings.

#43 Colorado

Colorado met 1 "yes" at large selection standard and 0 "no" standards.  Here is the "yes" standard it met:

RPI and Conference Rank (Standard #21)

The "yes" at large selection score for this standard is >=4.2184.

Colorado's RPI was .5914 and the Pac 12's Conference Rank was #1, combining to give Colorado a score for this standard of 4.2528.  If Colorado had not received an at large selection the required change for this standard would have caused 9 teams, of 404 unseeded at large selections over the last 12 years, that previously were in the "yes" at large selection category, to move out of it.

I have a comment about this particular standard.  The theory behind it is that it's increasingly difficult for a team, as its conference becomes stronger, to have a good RPI.  Thus as the Conference Rank gets better, less is expected of a team's RPI in order for it to get an at large selection.  The same is true for the standard that combines a team's Conference Standing and Conference Rank -- the better the Conference Rank, the less is expected of a team's Conference Standing.  There will be more related to this at the end of this article.

#57 Georgia

Georgia met 2 "yes" at large selection standards and 3 "no" standards.  This means it had a profile the Committee hasn't seen over the last 12 years.

Here are the "yes" standards Georgia met:

Top 50 Results Rank (Standard #6)

The "yes" at large selection score for this standard is <=12.0.

Georgia's Top 50 Results Rank was #11.  When I change the "yes" score for this standard to accommodate Georgia's not getting an at large selection, there will be 20 teams, of 404 unseeded at large selections over the last 12 years, that previously were in the "yes" at large selection category, that now will move out of it.

Top 50 Results Rank and Conference Rank (Standard #66)

The "yes" at large selection score for this standard is <=17.8.

The SEC's Conference Rank was #3, which combined with Georgia's Top 50 Results Rank to give it a score for this standard of 17.3.  With Georgia not getting an at large selection, there will be 2 teams, of 404, that previously were in the "yes" at large selection category, that now will move out of it.

Here are the "no" at large selection standards Georgia met:

RPI and Non-Conference RPI Rank (Standard #16)

The "no" at large selection score for this standard is <=3.2732.

Georgia's RPI was .5712 and its Non-Conference RPI Rank was #64, combining for a score for this standard of 3.2712.  If Georgia had gotten an at large selection the required change for this standard would have caused 3 teams, of 133, that previously were in the "no" at large selection category, to move out of it.

RPI and Common Opponents Score (Standard #23)

The "no" at large selection score for this standard is <=63.5213.

Georgia's Common Opponents Score was -5.63 (#59), which combined with its RPI to give it a score for this standard of 63.4783.  If Georgia had gotten an at large selection the required change for this standard would have caused 1 team, of 133, that previously was in the "no" at large selection category, to move out of it.

RPI Rank and Common Opponents Rank (Standard #35)

The "no" at large selection score for this standard is >=116.0.

Georgia's RPI Rank was #57 and its Common Opponents Rank was #59, combining for a score for this standard of 116.0.  If Georgia had gotten an at large selection the required change for this standard would have caused 2 teams, of 133, that previously were in the "no" at large selection category, to move out of it.

Comments on Georgia

Both of the "yes" at large selection standards Georgia met included its Top 50 Results Rank.  The "no" standards it met all included its RPI or RPI Rank, which were poor, and its Non-Conference RPI Rank or its Common Opponents Score or Rank, which likewise were poor.  It seems clear the Committee felt Georgia's Top 50 Results weren't good enough to overcome the poor other aspects of its profile.

Utah

Utah met 0 "yes" at large selection standards and 1 "no" standard.  Here is the "no" standard it met:

Head to Head Score and Common Opponents Score (Standard #86)

The "no" at large selection score for this standard is <=-11.6100.

Utah's Head to Head Score was -1.17 and its Common Opponents Score was -6.0 (#60), combining to give it a score for this standard of -12.1833.  With Utah getting an at large selection the required change for this standard will cause 7 teams, of 133, that previously were in the "no" at large selection category, to move out of it.

Overall Comments About the Committee's At Large Decisions

The Committee gave at large selections to Iowa and Utah over Florida Atlantic and Tennessee.  Although I've discussed why the Committee might have decided against Florida Atlantic and Tennessee, why pick Iowa and Utah?

Ordinarily, my first look is at teams' Top 50 Results Scores.  With Tennessee and Georgia having the best Top 50 Results Scores of the contending teams, that obviously wasn't the deciding factor so far as they were concerned.  Rather, other negative aspects of their profiles must have negated their Top 50 Results.  What about the other teams?

Here are the Top 50 Results Scores of the contending teams other than Tennessee and Georgia, including the 2 Ivy League Top 60 teams in contention that did not get at large selections and leaving out Yale which met 1 "yes" but 7 "no" standards:

Texas  4,680
TCU  3,470
Iowa  594
Colorado  570
Utah  549
Alabama  146
Harvard  20
Columbia  17
Florida Atlantic 2

A simple explanation for the Committee's decision is that after deciding "no" for Tennessee and Georgia, in making its last at large selections, the Committee looked to teams' Top 50 Results a the deciding factor.  Perhaps that's the right explanation.

There's also another possible explanation.  It has to do with Conference Ranks and Conference Standings.  The following lists, by conference, will show what I mean.  It goes through the Top 10 conferences according to the RPI, in order.  For each conference, the regular season champion and the conference tournament champion are in quotation marks ("").  The teams given seeds are marked with an asterisk (*).  To the left is the team's Conference Standing (average of regular season standing and conference tournament finishing position).  Unless specifically noted, all of the teams on the list that weren't the Automatic Qualifier got at large selections:

#1 Pac 12

"1 Stanford*"
2 UCLA*
3.5 Southern California*
3.5 Washington*
5 California
6.5 Washington State
6.5 Arizona
8 Utah
9.5 Colorado

9.5 Oregon State, did not get an at large selection.  Top 50 Results Score of 22.

[All other Pac 12 teams were below these in the standings.]

#2 ACC

"1 North Carolina*"
2.5 Virginia*
2.75 Florida State*
4.25 NC State
5.25 Louisville
6.5 Duke
6.5 Clemson
7.5 Notre Dame
8.5 Virginia Tech

[All other ACC teams were below these in the standings.]

#3 SEC

"1.5 South Carolina*"
"1.5 Arkansas*"
4 Florida
4 Vanderbilt
4.75 Texas A&M

6.25 Georgia, did not get an at large selection.
6.75 Albama, did not get an at large selection

[All other SEC teams, including Tennessee, were below these in the standings.]

#4 Big 12

"2.25 Oklahoma State*"
2.75 Texas Tech*
"3 Kansas*"
4 TCU
4.75 Texas
5.25 West Virginia

[All other Big 12 teams were below these in the standings.]

#5 Ivy League

"1 Brown"

#37 Yale, did not get an at large selection
#41 Harvard, did not get an at large selection
#49 Columbia, did not get an at large selection

#6 Big 10

2.25 Michigan
"2.5 Penn State*"
3 Rutgers*
"3.75 Wisconsin*"
5.75 Iowa

[All other Big 10 teams were below these in the standings.]

#7 West Coast

"1 BYU*"
2 Santa Clara
3 Pepperdine

[All other West Coast teams were below these in the standings.]

#8 Big East

"1 Xavier"
2 Georgetown

[All other Big East teams were below these in the standings.]

#9 American

"1.5 South Florida"
"1.5 Memphis"

[All other American teams were below these in the standings.]

#10 Conference USA

"1.5 North Texas" -- Automatic Qualifier

"1.5 Florida Atlantic," did not get an at large selection

[All other CUSA teams were below these in the standings.]

Looking at all of these together, they suggest to me two things:

(1)  The Committee paid a good deal of attention to conference RPI ranks.  The one exception is the Ivy League, for which I explained, in my first article on this year's Committee decisions, why the Committee might have concluded its RPI rank was wrong.  And, specifically for conferences, the Committee appeared to believe the Pac 12 and ACC were about equal; the SEC, Big 12, and Big 10 were about equal; they were followed by the West Coast (slightly ahead), Big East, and American; and Conference USA was a level below.

(2)  The Committee payed close attention to where teams within these conferences finished in their Conference Standings, including the regular season competiton and the conference tournament.

This seems to have had a couple of results:

First, for the #1 Pac 12 and #2 ACC, the Committee was willing to go deeper into their Conference Standings with at large selections.  In addition, for the ACC it was pretty clear that 8 of its teams should get at large selections, to go with its Automatic Qualifier.  With the Pac 12 being the better ranked conference, this may have caused the Committee to conclude that the Pac 12 teams also should get 8 at large selections.  This is particularly true since it would have been difficult to argue that the Pac 12 was the better ranked conference because it was stronger at the top -- not when the Committee gave the ACC 3 of the 4 #1 seeds.

Second, for the #3 SEC, #4 Big 12, and #6 Big Ten, the Committee wasn't willing to go as deep into their Conference Standings with at large selections, seeming to not want to give selections to teams with a Conference Standing of #6 or poorer.  For the SEC, only 5 teams were inside this limit; for the Big 12, there were 6; and for the Big 10, there were 5.  This may explain the distribution of at large selections among these three conferences.

Third, for the #7 West Coast, #8 Big East, and #9 American, their at large selections all were quite clear selections and they didn't have other teams within an RPI range to reasonably be considered for at large selections.

Fourth, Conference USA, especially after breaking down Florida Atlantic's profile, simply wasn't at a high enough conference level to justify it getting an at large selection.

Fifth, looking at the list makes it about as clear as is possible, that the Committee didn't buy the Ivy League's #5 RPI rank.

Finally, the Committee appears to have been unwilling, in general, to take teams out of their Conference Standing order.  Thus with Colorado as a likely selection in the Pac 12, this would have brought Utah in since it finished ahead of Colorado in the Pac 12 standings.

Whatever the individual Committee members actually were thinking, these explanations have satisfied me that the Committee's at large selection decisions were reasonable.