Which factors can indicate and give foresight to major first-round upsets in the Men’s NCAA Tournament?

By Dean Lowery '27

Introduction

The Men’s NCAA Tournament has a reputation for being unpredictable. Every year, millions of Americans attempt to fill out a perfect bracket in hopes of correctly predicting all 63 games that are contained in ‘March Madness’. However, some of this madness stands out above the rest: the ‘bracket buster’: the upset that comes as a surprise to seemingly every expert, breaking the hearts of millions of Americans and their dreams of a perfect bracket.

This investigation attempts to look for commonalities between these so-called bracket busters and how regular season data might help reveal red flags when millions of hopefuls are filling out their brackets each March. This research questions what statistics could indicate these forthcoming upsets, both for teams primed to upset and those likely to fall from their high-seeded throne.

Methods

For the purposes of this investigation, bracket busters are defined as any upsets of a one, two or three seed in the first round of the NCAA Tournament. This data sample consists of the last twenty of these upsets, beginning with 1999’s North Carolina v. Weber State and spanning through 2023’s Purdue v. Fairleigh Dickinson.

A confounding variable that could potentially distort results is the adjustment of the shot clock from 35 seconds to 30, implemented in 2015. While this change may first seem influential, within the scope of this research, the change is negligible; the statistics of note – namely shooting percentage - would be unaffected by change in pace, and statistics such as points per game have seen no significant impact since the 2015 change.

This research divided the teams from each of the twenty bracket busters into two categories: ‘losers’ and ‘winners’. The teams who emerged victorious and managed to upset are marked as winners, and those whose seasons perhaps ended prematurely were labeled as losers. Then, both the winners’ and losers’ team stats were organized and averaged. Thus, for each individual stat, the mean was calculated. Then, the standard variation of each of these statistics was taken and divided by its respective mean again, then multiplied by 100 for viewing convenience. This standard variation is a
metric that describes the spread of the data, or the average distance from each individual data point to the mean. This metric will indicate how similar each set of teams is for each statistic. This final calculated statistic will be referred to as an ‘importance score’, indicating how relevant this statistic is in finding commonalities in winners and losers of bracket busters.

For reference, a lower importance score indicates a higher correlation between that statistic and that category of team.

Results

Figures 1 and 2 below show the importance scores for winners and losers, respectively.

 

Figure 1: Importance Scores for Winners

Figure 2: Importance Scores for Losers

For winners of bracket busters, field goal percentage, two-point percentage, and three-point percentage serve as the three common factors. On the other hand, for losers, free throw percentage, field goal percentage and field goals attempted proved important. However, these figures seemingly are irrelevant to the argument, as these most important figures are often found at the middle of the pack for the winning and losing teams. In essence, the most common statistics between each set of teams are often around the median of all D-1 teams. There might be more significance if, for example, in one statistic the winners ranked poorly among D-1 teams. Then, as a conclusion, examine the Case Study of 2023’s Purdue v. Fairleigh Dickinson:

Relevant Statistic	Purdue Boilermakers	Fairleigh Dickinson Knights
Free Throw %	.744 (71st in D-1)	.752 (49th in D-1)
Field Goal %	.457 (105th in D-1)
Field Goal Attempts	55.3 (296th in D-1)
Two-Point %		.517 (134th in D-1)
Three-Point %		.345 (167th in D-1)

All these statistics are positive, such that being relatively better at them would indicate success, yet FDU was middle of the pack (out of 343 D-1 men’s teams) in both two-point and three-point percentages.

Discussion

So, what do these results indicate? Not much. While certain statistics emerged as more relevant than others, they don’t appear to be particularly helpful in identifying candidates to create bracket busters. These statistics would be relevant, for example, if the winners had relatively high statistics - for example in the top 10 in free throw percentage, given that they would be outstanding from other 14, 15 and 16-seeded teams. One statistic which appears to be relevant however is that the losers are often relatively worse in field goals attempted, meaning that they play at a slower pace, taking fewer shots and maintaining less possession. This appears to be relevant and logical, as a slower pace allows for more chances for hot shooting or a few loose balls to play a larger role in the outcomes of games. For example, a run of consecutive three-pointers made can have a much larger effect in a slow-paced game than in a faster, higher-scoring game.

Perhaps a logical interpretation of the findings is that a combination of a high-seeded, low-pace team with a low-seeded, hot-shooting team produces an upset. ‘Lucky’ refers not to good bounces or officiating fortunes, but rather an outstanding shooting performance that does not represent the shooting percentages of a team across an entire season. In the Purdue v. FDU example, Purdue ranked 296th in field goals attempted and ran into an FDU team that shot .302 from beyond the arc that night
compared to the Boilermakers’ .192.

Further exploration could be taken in how recent playstyle changes (a shift in preference towards three-pointers attempted) could reflect in more bracket-busting upsets per year.

Conclusion

Bracket-busting upsets seem to confuse the average fan and the dedicated expert it seems every year when it comes to the NCAA Men’s Basketball Tournament. However, a few key statistics seem to be tied to the winners and losers of these upsets. It appears that slow-paced teams are more prone to upsets, and when they happen to run into an outlier shooting performance, they tend to fall into the history books as disappointments. As the game of basketball continues to evolve, one thing remains
certain: these upsets aren’t going anywhere.

Data Sources

basketball-reference.com