Every year, millions of college basketball fans across the U.S. take to the internet to fill out their own copy of the NCAA Men’s Basketball Tournament bracket. Each person moves through the 68-team bracket game-by-game, round-by-round, choosing when they believe teams will ultimately return home. Some take the approach of choosing the highest seed of each match-up, while some choose to reflect on the teams’ records and trends in an effort to predict a few upsets along the road.
An upset is defined as a matchup in which the winning team is two or more seeds “lower” than the “higher” seed (i.e. an 11 seed beats a 7 seed, an 11 seed beats a 3 seed, etc.). These upsets are exactly where the madness comes into this March tournament and leads many to heavily analyze previous trends of upsets in an effort to potentially predict them in future tournaments.
However, in some tournaments, it seems nearly impossible, to many fans, to predict the many upsets that unfold. Fans should not get down on themselves when their seemingly perfect bracket is “broken” by an entirely unforeseen upset, though, because the chances of choosing the exact outcome and predicting those upsets for every single game in the tournament are very low. In 1985, as the NCAA points out, the tournament was converted from 53 to 64 teams (or 68, if including the “First Four” games immediately prior to full bracket play), so it is understandable that this is even less of a probable outcome in modern play. (For reference, according to the NCAA, no one has ever compiled an entirely correct March Madness bracket in the modern era and the odds of doing so are actually, statistically, 1 in 9.2 quintillion.)
In some years, fans feel extremely fooled by the sheer amount of upsets unfolding in the tournament. In particular, this past year, the 2018 NCAA Tournament, had fans everywhere bewildered. As mentioned before, an upset can be when an 11 seed beats a 7 seed, or a 3 seed, and so on. Last year, the 11-seeded Loyola University Chicago Ramblers did just that, along with also topping 6, 7, and 9 seeds to take them to the “Final Four,” the National Semifinals. Even more shocking to viewers was the fact that, additionally, for the first time in the history of the tournament, as USA Today reports, a 16 seed beat a number 1 seed. The University of Maryland-Baltimore County (UMBC) defeated the University of Virginia 74-54 in one of the first 28 games of the tournament, effectively breaking the 25 “perfect” brackets remaining at the start of the game and further tarnishing already-broken brackets across the nation, the NCAA states.
Fans of teams that became the victims of the many unlikely upsets in 2018, fought criticism saying things like: “Upsets are way more common now-a-days,” and “Look at how many other upsets there were this year!” It is no surprise that a year full of shocking endings of this magnitude began to stir talk of the tournament having significantly more and worse upsets than a “normal” year. But is this really the case?
To further analyze this question, it is necessary to look at the results from every NCAA Tournament game in the modern era, which includes data from 1985 to 2018, and determine which of these 2,189 games can actually be qualified as “upsets.”
How Common are Upsets?
The graph above depicts the number of upsets in each year of the tournament from 1985 to 2018. It can be seen that the number of upsets occurring varies greatly from year to year. For example, in 2006, 17 upsets took place, yet, the very next year, in 2007, there were only 4 total upsets throughout the entire tournament, an all-time low in the modern period. There does not appear to be any true trend throughout time. Each year brings different numbers of upsets and does not appear to hover around any given value nor does the number of upsets consistently increase or decrease across this time frame.
History of Upsets Compared to Average Number of Upsets
The dashed line represents the average number of upsets in the NCAA Tournament from this compiled group of games, working out to be approximately between 12 to 13 games (12.85294, to be exact). Therefore, it is obvious that in many years, the number of upsets falls well below this average. Likewise, many tournaments hit roughly at this point of containing about 12 to 14 upsets through the course of March Madness. This also depicts how some years, though, particularly stand out for their large amounts of upsets, like 2014, seeing a historic 19 upsets.
However, to further understand if the 2018 NCAA Tournament truly saw a historically large number of unsuspected outcomes, it is helpful to think about what percentage of games typically turn out as upsets. As seen above, the average ratio of upsets to all other game results is 20:80, meaning that, on average, around 80% of games actually result in what most people would expect: a higher seed (or a seed lower by less than 2 seeds) winning.
So, we can look at this average proportion and compare it to the tournament years found to be historically noteworthy in its amount of upsets, 2007 and 2014. This second figure shows that 2007’s minuscule 4 upsets ultimately reveals that only 6.35% of games seemed to be the start of everyone’s favorite “Cinderella Story” and the year saw 13.65% less upsets than average.
On the other side of the spectrum, filling out a close-to-perfect bracket 2014 took more unlikely prediction than a usual tournament with 28.36% of all games ending in upsets, 8.36% more than average.
Finally, it comes to last season’s whirlwind tournament. Technically, yes, at 22.39% of all games, the 2018 NCAA Tournament did contain more upsets than an average tournament. However, this is only 2.39% more upsets than a typical season and still 5.97% less than most upset-heavy tournament in the modern era. So, it is clear that the year was not necessarily record-breaking in its sheer amount of upsets, even though it did manage to break through the ceiling of average proportion of upsets.
This debunks the idea that the 2018 tournament presented significantly more upsets than a normal year, but what about the magnitude of the upsets?
How Big are the Upsets?
Visit the link here to use an interactive version of this graph!
(Use the legend to the right to toggle the levels of data viewable at once.)
The graph above shows the seeding differential (winning seed number – losing seed number) of each upset from 1985 to 2018. For example, no. 11 Loyola University Chicago’s win over no. 3 Tennessee would have a seeding differential of 8. So, the higher the value for seeding differential, the bigger, and possibly more shocking, the upset. A majority of the points are settled towards the bottom of the graph. This consequently portrays that most upsets seen in the March Madness Tournament are not large-scale upsets like that of no. 16 UMBC’s win against no. 1 Virginia, which is demonstrated as the point in the far top-right with a seeding differential of 15. Additionally, upsets historically occur less and less as the tournament progresses through its six rounds. When the information is isolated to only the Elite Eight, National Semifinals, and National Championship, there are considerably less points on the graph and they are more dramatically congregated around lower seeding differentials. This suggests that there are either less opportunities for upsets in these rounds because less low seeded teams have made it this far into the tournament, and/or the seeding match-ups simply become closer within these rounds, making it mathematically impossible for a win to be classified as an upset.
If we look at the 2018 year, we see that, despite the fact that the UMBC-Virginia upset was substantially larger than most upsets, it is actually one of few “big” upsets in that specific tournament. Following the trends of previous years, the 2018 tournament’s upsets seem to be settled further towards the lower and middle half of the seed differential scale, which is the norm for most of the data. However, when focusing on the upsets in the Sweet Sixteen, it appears that 2018 did in fact see more definable upsets in this one round than in most years. This can occur when the lower-seeded teams causing upsets in earlier rounds continue to win and therefore provide more opportunities for this trend of upsets to continue further into the tournament than fans have seen historically.
Although it does not seem that this past year’s March Madness had considerably more “large” upsets than in previous years of the modern tournament era, the upsets did exist deeper into the bracket than fans are used to seeing. This may explain why so many viewers felt as if they were witnessing a historic tournament in terms of upsets, as they saw lower-seeded teams weave their way further and further into March Madness.
Ultimately, there truly does not appear to be an increase and neither the amount of upsets nor the magnitude of upsets in the NCAA Tournament. The many outcomes the 2018 tournament, while still as surprising as you may imagine, actually do not stray exponentially far from the patterns of the tournament from the last 33 years, meaning that we cannot assume that upsets are becoming more common today.
So, happily enough for bracket-makers everywhere, the yearly outlandish guesses must continue. The Madness of March 2018 may not have made these predictions any easier for future tournaments, but there is no doubt that the upsets it brought helped keep fans on the edge of their seats far into the bracket. Their quest for a perfect bracket will forge on, but so will many teams’ missions to become the one team that “upsets” their way into breaking every single one of those brackets. And, from the looks of the historic disorder in the pattern of upsets, who knows, maybe one of these teams will help us see the first upset in the National Championship in 21 years someday soon.
*Original data from: Data World
*Featured photo: Picture: By Phil Roeder from Des Moines, IA, USA – NCAA Basketball, CC BY 2.0