top of page
  • Black Facebook Icon
  • Black Twitter Icon
  • Black YouTube Icon

The Perfect Bracket Project: Upsets, part One

jweimer25

Upsets are as inseparable from the NCAA Tournament as Reese’s cups are from my digestive tract. Trust me, at any given moment there’s at least one glorious amalgam of peanut butter and chocolate somewhere in my convoluted innards. If Reese’s University existed when I was in my teens, the Penn State diploma collecting dust in the cedar chest in my closet would never have made its home there.

 

All upsets, however, aren’t created equal. A 10-seed toppling a 7 on a Thursday afternoon might count as an upset, but it will never shake up the world like the Fairleigh Dickinson Fightin’ Mini-Me’s chopping down the Zach Edey goliath. If we're going to have a crack at a perfect bracket, we need some thoughtful analysis of how many upsets to expect, what kind, and when. This is truly the first layer of the onion that we need to peel.

 

Let’s start with the big picture: defining an upset. Different sources will yield different definitions--some basing it on the closing betting line, some on seeding, some on team strengths according to a comprehensive source like KenPom or ESPN's BPI. This site will define it strictly by seeding: an upset is any occasion when a higher-seeded team defeats a lower-seeded one, regardless of what other metrics say.


Why this definition, you might ask? I'll tell you. The NCAA Selection Committee has one job: selecting and seeding the 68 teams that make up the tournament field. They spend countless hours watching games, poring over resumes, evaluating one team versus another. They were chosen specifically for the task. So to define an upset by any other metric is to cast a vote of no confidence in the work the Committee has been entrusted to do. Many people do this, of course, and there may be good reasons to do so. But I'm a trusting guy. I choose to believe the Committee knows what they're doing.


Also, basing upsets upon seeding makes the research side of this project a whole lot easier.


With that baseline settled, we can peel down to the next layer: how many upsets happen annually. Conveniently, we have a chart:



As this graphic displays, in spite of the fact that the NCAA Tournament is an annual cavalcade of surprise, the total number of upsets per year is remarkably consistent. The tournament average is just below 18 total upsets, and 27 of the 39 tournaments contested in the 64+ team era have featured between 16 and 20 (69%). The high bar is 23 upsets, in 1992, and three events hit the low bar of 12 (1993, 2007, and 2015). Through this we can start to sketch out my first Perfect Bracket Parameter: never pick fewer than 12 total upsets, and aim for that sweet spot of 16 to 20.


It isn't that simple, however. There are, as we mentioned above, different types of upsets. Some upsets may not even be regarded as such by the sharps in Las Vegas (plenty of 9, 10, and 11 seeds close as the betting favorite, for example).


So, for our analysis, we’re going to separate upsets into 3 buckets:

 

1.      Minor Upset: Any upset in which the difference between the two teams’ seeds is less than 5. This covers the 8-9 and 7-10 games in the first round, and occasions like a 6 over a 3 or a 3 over a 2 in the tournament’s later stages. This is that "gray zone" where an upset may be such on paper, but whether it's truly an "upset" can be up for debate.

2.     Major Upset: Any upset in which the seed difference between teams is 5 or greater. From the annual rite of passage known as the 5-12 upset, to the 1 and 2 seeds bowing out before the sweet 16, we know these upsets will happen, but we're still mostly shocked when they do.

3.     Bracket Buster: Even in the range of major upsets, there are outliers. These are the upsets that get replayed on highlight montages year after year; the ones that diehards can remember by a mere mention of a small school name: UMBC, St. Peter’s, George Mason, Florida-Gulf Coast. The term bracket buster is probably overused, but many upsets definitely fit the moniker. I define a bracket buster as any upset that occurs when the teams' seed difference is 10 or more. Any 14-, 15-, or 16- seed winning in the first round counts here, as do those rare instances when an 11 beats a 1 (see: George Mason, VCU), a 15 beats a 3 (St. Peter’s over Purdue, in the sweet sixteen, after beating 2-seed Kentucky in round one in perhaps the most remarkable run the tournament has ever seen), and so on.


This next chart shows how these 3 upset categories are segmented across the 39 tournaments in the 64+ team era:



Let’s parse this out a bit. Over the course of 39 tournaments, we’ve witnessed 695 total upsets—an average of just under 18 per year. Just over half of these (370, or 53.24% to be exact) fall into the minor upset category.

 

Major upsets, or those in which the seed difference is between 5 and 9, comprise 40.86%—284 in total, or 7.28 per tournament. This number is rather notable. The average NCAA Men’s Tournament features between 7 and 8 upsets in which the seed difference is greater than 5—the 5-12 and 4-13 shockers, the double digit seed appearances in the sweet 16, the surprising runs by 7- or 8- seeds. These are the upsets that shape the contours of a given tournament’s story, the ones that befuddle experts and have derailed all attempts at a perfect bracket thus far. These (plus the bracket busters, which we’ll look at in a moment) are the ones to decipher, somehow, in our quixotic quest. We’ll get to that.


Also quite remarkable is the fact that a true bracket buster upset happens about once per tournament. We’ve seen 41 bracket busters on the whole. As a scan of the full list of these upsets[i] shows, the vast majority of these (36 of 41) happened in the first round. From Cleveland State to Georgia State to Oakland, a 14 defeating a 3 is the most common bracket buster, but 15-seeds have had their day versus 2’s on an increasing scale since the Richmond Spiders broke the glass on that ceiling against Syracuse in 1991. And, of course, UMBC and Fairleigh Dickinson have finally, memorably, given 16-seeds their day in the not-too-distant past.

 

The 5 “other” bracket buster upsets, the one’s outside the first round, are worth specific mention. Somewhat astonishingly, four of these five were instances of an 11-seed knocking off a 1-seed to reach the final four: LSU over Kentucky in 1986, George Mason over UConn in 2006, VCU over Kansas in 2011, and UCLA over Michigan in 2021. In fact, 11-seeds are an even 4-4 versus 1-seeds in the Elite 8. Of all the seed lines, 11’s are an outlier in terms of expected performance. They punch above their weight, and 39 tournaments are a large enough sample size to say that this isn’t random noise in the data. We’ll speculate why in a future post.


Finally, the one bracket buster that didn’t happen in either the first round or the Elite 8 was Saint Peter’s upset of Purdue in the 2022 Sweet 16. Part of this is due to probability--it’s mathematically impossible for a bracket buster upset to happen in the second round (the seed difference in this round can never be greater than 8), and the elite 8 and final 4 have a far greater range of possible seeding matchups than the sweet 16 does. Bracket buster possibilities can happen at this point in the tournament, but they require very specific scenarios: a 16-seed matching up with a 4 or 5 (hasn’t happened yet), a 15 matching up with a 3 (three times), a 14 meeting a 2 (never). The most common bracket buster opportunity at the sweet 16 level is when a 12 or 13 seed meets a 1. This has occurred 24 times, but the 1 has never lost. Yet.


Before we start to draw conclusions on all this, let’s look at one more interesting piece of data. If you’re as committed a tournament watcher as I am, you’ve probably developed an intuitive sense that the madness has only gotten madder in recent years. It feels like the upsets are wilder and more frequent, doesn’t it? At a bracket-buster level, consider that, since 2018 we’ve seen:

·      The heretofore impossible 16 over 1 upset twice (2018, 2023).

·      A 15-seed advancing to the sweet 16 in three consecutive years (2021-23).

·      The 15th-seeded Saint Peter’s Peacocks advancing to the elite 8 (2022).

 

But does the data align with this intuitive sense? Take a look:



This chart shows the cumulative average of each type of upset. Each point on the graph shows what the average was at that point in time. For example, the top line represents the total number of upsets in each tournament. The first 3 years—1985, 86, and 87—each featured 18 upsets, so the cumulative average held steady at 18. There were 17 upsets in 1988, moving the average at that point to 17.75. Each subsequent tournament added new data, and the averages changed accordingly.

 

Plotting these cumulative averages gives us visual trendlines for each upset type. As the first graph shows, the early 90's were not friendly to upset lovers, pulling the cumulative average down to a low of 16.69 upsets per tournament as of 1997. Since then, the average has risen by over 1 upset per event to its current 17.82 average. Looking deeper, the average over just the last 10 tournaments is 18.6, and even that number is being held down by the somewhat-abnormal 2015 and 2017 events, which featured just 12 and 14 upsets respectively. The other 8 tournaments held in the last 10 years (well, 11, given the blank COVID year) have all had 19 or more upsets.


As for the specific upset types, after early years that featured lots of major upsets and just a few minors, the cumulative averages for these leveled out by the mid-to-late 90's and have remained relatively steady. The cumulative average for minor upsets has even dipped slightly, from a high of 10.1 in 2005 to a current level of 9.49. That has been balanced by an uptick in major upsets, from a cumulative average of 6.36 in 1998 to the current average of 7.28.


Bracket busters are perhaps the most interesting case. After none in the initial 64-team event in 1985, the 1986 bracket exploded with 3 such upsets: wins by 14th-seeded Cleveland State and Arkansas-Little Rock, and the historic run by 11 seed LSU to the Final Four, featuring an Elite 8 win over top-seed Kentucky. This 1986 explosion was followed by a relative trickle of bracket buster upsets, drawing the cumulative average down to a low point of 0.84 per tournament by 2009. Since then we've seen a steady increase to the current 1.09 average. This might seem non-noteworthy, but is worth a deeper look.


Our era consists of 39 tournaments: 25 from 1985-2009 and 14 since. Nearly half of all bracket buster upsets (20 of 41) have happened between 2010 and now. The per-tournament average in this stretch is nearly 1.5 per tournament. Given that most of these upsets happen in the first round, this means that a 14, 15, or 16 seed is breaking through at least once per tournament. We even had another explosion of 3 bracket busters in one year, in the Indiana-bound partial-lockdown 2021 tournament (14-seed Abilene Christian over Texas and 15-seed Oral Roberts over Ohio State in the first round, and another 11 over 1 takedown in the Elite 8, UCLA over Michigan). Growing up, and even in recent years, it was a safe bet for me simply to pencil the 1, 2, and 3 seeds into the second round. Upsets would happen, but not in the more-than-one-per-year frequency we see now. Picking a perfect bracket requires one to hit on these rarities, quite possible the tallest task in our quest.


All of this proves that, yes, the madness is indeed getting madder. Upsets are happening more frequently (even if just a little bit), and those upsets are on the more extreme side. Given the transfer portal, NIL era we currently find ourselves in, I don't think it will be slowing down. Parity is firmly entrenched in men's college basketball. If we want to capture our white whale of a perfect bracket, we'll need to be like Salt Bae and liberally sprinkle upsets across the bracket. This post helps us understand a bit more about how to divide that salt; the next will explore more of where to sprinkle it, by looking at what rounds are the most and least upset prone.



· [i]1986 First Rd: (14) Cleveland State over (3) Indiana

·      1986 First Rd: (14) Arkansas-LR over (3) Notre Dame

·      1986 Reg. Final: (11) LSU over (1) Kentucky

·      1987 First Rd: (14) Austin Peay over (3) Illinois

·      1988 First Rd: (14) Murray State over (3) NC State

·      1989 First Rd: (14) Siena over (3) Stanford

·      1990 First Rd: (14) Northern Iowa over (3) Missouri

·      1991 First Rd: (15) Richmond over (2) Syracuse

·      1991 First Rd: (14) Xavier over (3) Nebraska

·      1992 First Rd: (14) East Tennessee St. over (3) Arizona

·      1993 First Rd: (15) Santa Clara over (2) Arizona

·      1995 First Rd: (14) Old Dominion over (3) Villanova

·      1995 First Rd: (14) Weber State over (3) Michigan State

·      1997 First Rd: (15) Coppin State over (2) South Carolina

·      1997 First Rd: (14) Chattanooga over (3) Georgia

·      1998 First Rd: (14) Richmond over (3) South Carolina

·      1999 First Rd: (14) Weber State over (3) North Carolina

·      2001 First Rd: (15) Hampton over (2) Iowa State

·      2005 First Rd: (14) Bucknell over (3) Kansas

·      2006 First Rd: (14) Northwestern St. over (3) Iowa

·      2006 Reg. Final: (11) George Mason over (1) UConn

·      2010 First Rd: (14) Ohio over (3) Georgetown

·      2011 Reg. Final: (11) VCU over (1) Kansas

·      2012 First Rd: (15) Lehigh over (2) Duke

·      2012 First Rd: (15) Norfolk State over (2) Missouri

·      2013 First Rd: (14) Harvard over (3) New Mexico

·      2013 First Rd: (15) FGCU over (2) Georgetown

·      2014 First Rd: (14) Mercer over (3) Duke

·      2015 First Rd: (14) Georgia State over (3) Baylor

·      2015 First Rd: (14) UAB over (3) Iowa State

·      2016 First Rd: (14) Stephen F. Austin over (3) West Virginia

·      2016 First Rd: (15) Middle Tennessee over (2) Michigan State

·      2018 First Rd: (16) UMBC over (1) Virginia

·      2021 First Rd: (14) Abilene Christian over (3) Texas

·      2021 Reg. Final: (11) UCLA over (1) Michigan

·      2021 First Rd: (15) Oral Roberts over (2) Ohio State

·      2022 First Rd: (15) Saint Peter’s over (2) Kentucky

·      2022 Sweet 16: (15) Saint Peter’s over (3) Purdue

·      2023 First Rd: (15) Princeton over (2) Arizona

·      2023 First Rd: (16) Fairleigh Dickinson over (1) Purdue

·      2024 First Rd: (14) Oakland over (3) Kentucky

 

7 views0 comments

Recent Posts

See All

Introducing the Perfect Bracket Project

Ever since I started filling out NCAA tournament brackets (the halcyon years of the mid-90’s), I’ve dreamt about filling out a perfect...

留言


© 2024 by Jason Weimer. Powered and secured by Wix

bottom of page