How The Heck Did We Get This Final Four?


This article is part of our March Madness series.

We’re not exactly strangers to low-probability events here at FiveThirtyEight. But even for a month defined by uncertainty on the hardwood, this March dealt us an extra helping of disruptions to your scheduled programming. The first matchup of the men’s Final Four next Saturday will be between two first-time attendees, No. 5 seed San Diego State and No. 9 Florida Atlantic, in a game that our model gave a 0.05 percent chance (!) of happening before the tournament. Meanwhile, the nightcap between No. 5 Miami (also a Final Four newbie) and No. 4 UConn had an astronomically higher chance of happening — all the way up at 0.3 percent.15

How do we make sense of this seemingly senseless madness? Well, it helps to put things in a historical context. Since 2002, when Ken Pomeroy started tracking efficiency numbers in college basketball, only the 2011 tournament saw a less-heralded quartet make it to the final weekend of the tournament — both in terms of predictive metrics and where the teams were seeded:

The 2023 Final Four is among the least chalky

Average pre-tournament seed and adjusted offensive, defensive and overall efficiency ranks for men’s NCAA Tournament Final Fours, since 2002

Season Seed Offense Defense Efficiency
2023 5.8 28.5 49.0 21.0
2022 3.3 12.0 41.3 14.5
2021 3.8 9.5 39.0 13.8
2019 2.8 12.5 14.8 6.8
2018 4.0 26.0 24.3 15.5
2017 3.0 45.3 13.0 12.8
2016 3.8 25.5 15.5 13.8
2015 2.5 6.3 23.0 6.8
2014 4.5 24.5 24.5 14.3
2013 4.5 21.8 29.5 13.3
2012 2.3 37.0 4.5 6.8
2011 6.5 30.5 61.3 37.5
2010 3.3 25.0 17.8 12.0
2009 1.8 21.0 17.3 9.5
2008 1.0 8.5 7.0 2.5
2007 1.5 7.0 13.5 4.5
2006 4.5 32.0 16.0 13.3
2005 2.8 6.3 12.0 4.3
2004 2.0 10.3 9.0 4.8
2003 2.3 10.0 40.5 12.0
2002 2.3 16.3 7.3 5.3

Source: KenPom.com

But digging deeper into just how much this group upended our expectations, it’s more of a mixed bag. Prior to the start of March Madness, we usually look at the top teams that are most likely to suffer an early exit — and by the characteristics that ring alarm bells, Miami was setting off all the sirens and klaxons. Not only did the Hurricanes have an adjusted defensive ranking of just No. 132, the worst mark of any eventual Final Four team since 2002, but they also were exceptionally overseeded: By KenPom’s metrics, Miami looked more like a No. 10 seed than a No. 5, making them the second-most overseeded Final Four team in that period.

The metrics didn’t believe in the Canes …

Men’s NCAA Tournament Final Four teams that were overseeded by at least two lines according to Ken Pomeroy’s adjusted efficiency rankings, 2002-2023

Season Team Eff. RK Off. rk Def. rk KP seed* Act. Seed Over​seeded by
2011 VCU 82 60 126 21 11 10
2023 Miami 40 12 132 10 5 5
2011 Butler 44 34 69 11 8 3
2018 Kansas 9 6 46 3 1 2
2009 Michigan St. 14 35 9 4 2 2
2009 Villanova 17 28 20 5 3 2
2003 Syracuse 20 18 31 5 3 2
2003 Texas 9 3 48 3 1 2

*Refers to seed in hypothetical scenario where teams were seeded according to Ken Pomeroy’s pre-tournament metrics.

Source: kenpom.com

However, two of the other three Final Four teams are actually among the most underseeded teams in recent Final Four history. As you can see in the table below, Connecticut — the overwhelming favorite left in the tournament, at 43 percent in our model (for whatever that’s worth this year) — is on the opposite end of the spectrum from Miami, along with Florida Atlantic, which made it to the tournament’s final weekend during its first trip to the Big Dance in more than two decades.16 And though San Diego State didn’t make the cut for our list, the Aztecs were also underrated by a full seed line.

… but they loved the Huskies and Owls

Men’s NCAA Tournament Final Four teams that were underseeded by at least two lines according to Ken Pomeroy’s adjusted efficiency rankings, 2002-2023

Season Team Eff. rk Off. rk Def. rk KP seed* Act. Seed Under​seeded by
2006 George Mason 25 75 16 7 11 4
2023 UConn 4 6 18 1 4 3
2014 Kentucky 19 19 32 5 8 3
2005 Michigan St. 6 7 26 2 5 3
2002 Indiana 8 31 6 2 5 3
2023 Florida Atlantic 26 32 36 7 9 2
2015 Michigan St. 18 15 38 5 7 2
2005 Louisville 8 13 11 2 4 2

*Refers to seed in hypothetical scenario where teams were seeded according to Ken Pomeroy’s pre-tournament metrics.

Source: kenpom.com

All of that leaves us with a slightly complicated view of the Final Four field. We’d be lying if we said we saw these matchups coming (our model implied a 1-in-412,740 chance of this particular quartet playing in Houston), but it is fascinating that arguably the most unexpected entrant isn’t the No. 9 seed that most of us hadn’t heard of before two weeks ago, but rather the experienced, guard-heavy team that returned most of its contributors from last year’s Elite Eight run. Perhaps the Hurricanes are just a metrics-defying team — after all, they were ranked No. 62 headed into last year’s event, too. But it’s more likely that this is just an artifact of March’s sheer unpredictability. (You would have made a killing if you had bet on a Miami Final Four run when the Canes were down by 8 points to No. 12 seed Drake with five minutes left in the round of 64.)

And if 2011 (which also held the Final Four in Houston) is the only peer to the 2023 tournament in terms of pure randomness, we can only hope that this Big Dance doesn’t end like its historical predecessor: with an ugly championship-game rock fight, accompanied by Rebecca Black’s “Friday.”

Check out our latest March Madness predictions!