What Historical Analytics Data Tells Us About the 2019 NCAA Tournament Bracket
Never worry about busted brackets again! Sign up to play SI’s Realtime Bracket game. If your original pick is losing, you can switch it during live tournament games. Game leaderboards track original AND realtime picks.
Filling out an NCAA tournament bracket can mean different things for different people. Some simply pick their favorite team to win it all and call it a day; others act off a gut feeling based on whatever amount of college basketball they watched during the regular season, and still others scour the data in search of every edge they can find. If you’re one of the many people who don't spend more than 10 minutes filling out their bracket, this guide probably isn't for you. But if you're into stats, or just trying to learn about what makes teams more or less likely to win a national title or avoid a first-round upset, these four points of analytical guidance may help.
1. Profile of a Champ
Two years ago, former SI.com writer Luke Winn wrote a piece about the profile of a national champion, in which he singled out four key factors: elite offensive efficiency, pretty elite defensive efficiency, a relatively low reliance on three-pointers and strong interior defense.
Last year, Villanova partly strayed from that profile—the Wildcats were heavily reliant on threes, ranking 12th nationally in three-point rate and getting 40% of their scoring from the perimeter. They also had a merely average interior defense, finishing the year ranked 140th nationally in field goal percentage allowed on two-pointers. But ‘Nova’s offense was so historically strong that it didn’t matter, and its 41.1% mark from three across six tournament games helped carry it to the title, where its offense and sturdy-but-not-amazing defense won out over Michigan’s elite defense.
It’s unlikely we’ll see a champ like Villanova this year. None of the field’s No. 1 or No. 2 seeds rank in the top 100 in three-point rate, with Michigan having the highest at No. 132 (the Wolverines are average perimeter shooters at 35%). Lower-seeded teams like Purdue, Virginia Tech, Auburn, Villanova, Louisville and Wofford do all take a high number of threes, but any member of that group would have to overcome several big obstacles to win the title.
Instead, we’re likely to get a more traditional champion. Of the 10 teams with the highest odds of winning the title on FiveThirtyEight, here’s how they stack up to those four key factors on their kenpom national ranks:
A few things to keep in mind:
• Remember, that’s three-point rate, not percentage. Historically, teams that are less reliant on threes have a better chance of going deeper in a tournament setting. Purdue’s No. 40 ranking is the clear outlier here, and even Michigan’s feels a bit high. Villanova’s two recent title teams are the only ones since 2006 Florida with a three-point rate that finished ranked higher in the D-I rankings than the Wolverines’ current position at No. 132.
• Duke’s rankings took a bit of a hit in Zion Williamson’s absence. Its offense, defense and two-point D were all higher before his injury, and since he’s healthy again, it’s fair to bump all of those up a bit in your mind.
• There’s a mix of very balanced teams (Duke, Virginia, North Carolina, Michigan State, Kentucky) and more unbalanced ones at the top. Gonzaga, Tennessee and Purdue are reliant on their offenses, and the Zags and Vols both check the other three boxes of Winn’s champion profile outside of defense. Gonzaga is in a particularly strong position because it has the No. 1 offense—and did even before Williamson’s injury—and a top-20 defense, while Virginia being the only team that is top-five in both puts it in an excellent spot. Michigan and Texas Tech, meanwhile, are defensively driven. The Red Raiders have seen their offensive efficiency surge down the stretch, but if Texas Tech won the title, it would join 2014 UConn as the only team since 2002 that entered the Big Dance ranked outside the top 21 on offense on kenpom and won it all. Additionally, 12 of the last 17 champs entered the tourney ranked in the top 11 on offense.
• Finally, you’ll notice that outside of UNC and Purdue, all of these teams are among the 25 best in the nation at two-point defense. That’s not a coincidence.
2. Momentum: Myth or Indicator?
Momentum is one of the most hotly debated concepts in sports. Does playing well one day carry over to the next? What about playing well for a week? A month? Every year in college basketball, we see teams that head into the NCAA tournament on hot streaks naturally garner extra attention, while teams that struggled down the stretch are viewed as ripe for an upset. But there’s more that defines whether a team is “playing well” than just wins and losses.
T-rank, an analytics site similar to kenpom that keeps tempo-free stats, has helpful filters that let you (among other things) view its rankings through any particular window of dates. We looked at the rankings from Feb. 1 to March 13, 2018, or the month and a half window before last year's NCAA tournament began. The top three teams were Michigan, Virginia and Villanova—the two eventual national championship game participants … and one that lost to a No. 16 seed. The pure improbability of that historic loss makes it easy to consider it an outlier, but otherwise, two of the top three teams were the last two standing.
What about other years? We went back 10 seasons—from 2009 on—and looked at where the national championship game participants ranked in T-rank in the window from Feb. 1 to the start of the NCAA tournament. In the last four years, all eight national title game participants have landed in the top six in that timeframe, but things open up a bit in the years before that. In all, 12 of the 20 last championship game participants entered the NCAA tournament ranked in the top six down the stretch, while 14 were in the top 12. The biggest outliers were 2011 Butler (No. 48), 2010 Butler (No. 39) and 2013 Michigan (No. 34). All of those teams lost the title game—two against teams that ranked No. 1 in the same window.
So who does T-rank have as the top teams since Feb. 1 entering the 2019 NCAA tournament?
1. North Carolina
2. Texas Tech
3. Gonzaga
4. Michigan
5. Virginia
6. Kentucky
7. Duke*
8. Michigan State
9. Houston
10. Penn State (not in tournament)
11. Tennessee
12. Kansas State
13. Purdue
14. Florida State
*Duke played six games in this span without Williamson. Prior to Williamson’s injury, the Blue Devils ranked second for the season on T-rank.
3. Be Wary of Extreme Teams
Every season in the kenpom.com era (since 2002), without fail, there are a group of teams that I classify as “Extreme Teams.” I define an Extreme Team as one that ranks in the top 10 in efficiency on either offense or defense, and outside the top 50 in the other; it’s elite on one end of the floor and not so much on the other. I was curious what this type of profile has historically meant for a team’s Final Four chances, especially considering it was only two years ago when a South Carolina team that entered the tournament with the No. 3 defense and No. 149 offense made it all the way to the final weekend as a No. 7 seed.
Since 2002, there have been 141 Extreme Teams in college basketball. Of those, 122 (86.5%) made the NCAA tournament, but none won the national title. Thirteen (10.6%) of those 122 advanced to the Elite Eight, with only four (0.3%) making the Final Four. A fifth, 2013–14 Wisconsin, just missed out on qualifying as an Extreme after entering the tournament with the No. 50 defense.
Only one team, 2012–13 Michigan, which entered with the No. 2 offense and No. 66 defense, made the national title game. The Wolverines improved defensively during their tournament run, finishing at No. 37. Improvement over the course of the tournament is common by teams that succeed as an Extreme—for example, ’16–17 South Carolina’s offense moved up 58 spots to No. 91 at season’s end—but not always a requirement.
Success, however, is the operative word here. Plenty of Extreme Teams have flamed out early, some stymied by opponents that either exploited their weak end of the floor or neutralized their strength. Since we’re talking about the NCAA tournament, of course, it’s important to consider where a team was originally seeded in the bracket. We’ve established that it’s immensely unlikely to get to the Final Four as an Extreme, but how many teams were in a realistic position to get there in the first place?
First, here’s the seeding breakdown of those 122 teams:
No. 1 seeds: 3 teams
No. 2 seeds: 8 teams
No. 3 seeds: 13 teams
No. 4 seeds: 19 teams
No. 5 seeds: 16 teams
No. 6 seeds: 13 teams
No. 7 seeds: 11 teams
No. 8 seeds: 7 teams
No. 9 seeds: 6 teams
No. 10 seeds: 9 teams
No. 11 seeds: 9 teams
No. 12 seeds: 5 teams
No. 13 seeds: 2 teams
No. 14 seeds: 1 team
Viewing each of these seeding lines through the prism of whether these teams met or achieved their seeding expectations (i.e. did a No. 1 seed reach the Final Four? Did a No. 4 seed make the Sweet 16?) yields some interesting results. All of the No. 1 seed Extremes fell short, including Xavier, which was upset by Florida State in the second round last year. No. 2 seeds have also been extremely vulnerable: Of the eight, seven failed to get out of the first weekend, and three lost to No. 15 seeds in the first round. No. 3 seeds have fared decently better, while No. 4 seeds have done fairly well—11 of the 19 teams met or exceeded their seeding expectations (i.e. reached the Sweet 16).
However, in an interesting twist, there’s a major gap in how No. 4 and No. 5 seeds have performed. The 5–12 game is known for being ripe for an upset, and that’s been especially true when it comes to Extreme Teams. Eleven of the 16 fifth-seeded Extremes lost to the No. 12 seed in the first round, and only three got to the Sweet 16 (none further). A whopping 41% (11 of 27) of all the 5–12 upsets that have happened since 2002 came at the expense of an Extreme Team.
Below, you can see a chart showing the percentage difference for every historical seed line in terms of how far a team with that seed is expected to advance (found via Bracket Odds) vs. how Extreme Teams with that seeding have done since 2002. Green means those Extreme Teams, on average, have done better than their seeds’ Bracket Odds percentage; red means they’ve done worse.
Now that we’ve established all that, what does all this tell us about the 2019 field? This year, there are five Extreme Teams in the tournament, but we’re going to go ahead and call it six, since Wofford is right on the cusp with the No. 12 offense (only a fraction behind No. 10 in efficiency):
Of these six, Kansas State and VCU are “Super Extremes,” or what I call teams that are in the top 10 on one end and outside the top 100 in the other.
There are no No. 1 or No. 2 seed Extreme Teams this year, but there is a No. 3 (LSU), No. 4 (Kansas State), No. 5 (Wisconsin), No. 6 (Iowa State), No. 7 (Wofford) and No. 8 (VCU). If you followed along with the above, you’ve probably already figured out that based on historical precedent, Wisconsin seems to be in the most danger here. While the Badgers only barely qualify as an Extreme, their offense has been trending in the wrong direction, going from No. 25 on Feb. 5 to No. 52 now. In their favor, however, is the fact that their opponent, Oregon, is far from an offensive juggernaut itself, and a potential second-round matchup with fellow Extreme Kansas State could await.
There’s something else we can especially glean from that chart: No. 7 seeds have been the least bothered by being an Extreme Team, which bodes well for Wofford. Five of the 11 No. 7 Extremes exceeded expectations by making the Sweet 16, including 2017 South Carolina making the Final Four. Four of those knocked off No. 2 seeds, while two lost to the No. 10 in the first round. The difference could perhaps be due to the fact that for any No. 7 seed to overachieve in March Madness, it typically needs to do something important particularly well. Being top-10 nationally on one end of the floor certainly qualifies.
4. Beware This Red Flag
When a team turns the ball over, it loses out on a potential shot, often ending a possession before even attempting one. The more turnovers, the more potential shots a team is missing out on, to say nothing of the transition opportunities giveaways tend to produce for opponents.
It’s no surprise then that this area can historically correlate to a level of March success. To be clear, a team doesn’t need to be elite at taking care of the ball (though it certainly is an advantage!), but if it has Final Four aspirations, being poor at it could be a significant hindrance.
Of the 68 Final Four teams since 2002, 47 (69%) finished their season (including the NCAA tournament) with an offensive turnover rate ranked inside the top 100. Twenty-nine teams (43%) were especially strong in that area, finishing in the top 50. Only five of the 68 Final Four teams (7.4%) finished outside the top 200.
Let’s look at where 2019 tournament teams currently rank in offensive turnover rate on kenpom. It should be noted that while these are pre-tournament numbers (as compared to post-tournament numbers used above) a few games aren’t likely to significantly alter a full season’s rate, and turnovers are not an area that is likely to be magically fixed in the tourney. However, a top-100 finish is especially within reach for those sitting on the fringe right now.
First, here’s a listing of this year's best teams in offensive turnover rate heading into the tournament. These are the teams currently within the top 50: Michigan, St. John’s, Nevada, Wisconsin, Gonzaga, Virginia, Belmont, Buffalo, Temple, North Dakota State, Tennessee, Purdue, Cincinnati, Iowa State, Wofford and Saint Mary’s.
Next, let’s look at the teams ranked from 100 to 199, with teams that are a top-five seed in their region in bold:
Oklahoma (100), Seton Hall (102), Montana (103), Auburn (106), UCF (107), Virginia Tech (109), Minnesota (113), Northern Kentucky (115), New Mexico State (117), Louisville (120), Duke (122), Oregon (125), Utah State (126), Old Dominion (125), Florida (136), Yale (139), Texas Tech (148), Ole Miss (151), LSU (155), Syracuse (169), Arizona State (179), Kentucky (180), Michigan State (188), Ohio State (189), Kansas (195).
You’ll notice that list (and the below one) is heavy on East Region teams. You have to go all the way down to No. 11 seed Belmont/Temple to find an East team currently within the top 100, so no high seed in that corner of the bracket has a particular advantage in the turnovers department.
Finally, let’s look at teams with a bright red flag—those that are currently outside the top 200 in offensive turnover rate. Once again, top-four seeds are in bold:
Bradley (209), Saint Louis (214), Florida State (225), Mississippi State (227), Colgate (232), Fairleigh Dickinson (237), Marquette (240), Washington (258), Maryland (262), Baylor (264), VCU (280), NC Central (343).
The main thing to learn from that? Take a long, hard look before picking Florida State, Mississippi State, Marquette or Maryland as a sleeper Final Four team. Unless they suddenly start taking better care of the ball in the tournament, history says a run to Minneapolis will be highly unlikely—but not unprecedented (for one thing, Loyola-Chicago did it just last year, but it was a clear outlier from its co-Final Four teams).
Meanwhile, every national title winner in the kenpom.com era (since 2002) has finished the year ranked inside the top 200 in offensive turnover rate, and in the last 10 seasons, the champ has finished the tournament either in the top 100 or right outside of it (2014 UConn finished 103rd). That’s not encouraging news for Texas Tech, LSU, Kentucky, Michigan State and Kansas (Duke, currently at No. 122, is in a better position to crash the top 100).
Nothing is guaranteed in March, and as we learned last year, nothing is off the table. But we can still learn from the past to get a sense for what's coming next.
