How to Win Your March Madness Bracket (or at Least Get Closer)
Every March, millions of people fill out NCAA Tournament brackets with a mix of confidence, chaos and instinct. Some rely on seeding, others trust their favorite teams and most are simply hoping to beat their friends. But if you want to build a better March Madness bracket, Creighton University math professor Nathan Pennington, PhD, says it starts by rethinking how the bracket actually works.
In Creighton’s latest Fast Class, Pennington breaks down how probability, analytics and context shape the outcomes we try so hard to predict. While the tournament is known for its unpredictability, his approach reveals that there are patterns behind the chaos if you know where to look. It starts with a simple idea: the number next to a team’s name does not always tell the full story.
Looking beyond the seed line
It’s easy to assume that NCAA Tournament seeding is a clean ranking of the best teams. A No. 1 seed should be better than a No. 2, and a No. 5 should be stronger than a No. 12. But that’s not exactly how the bracket is built. As Pennington explains, “The NCAA selection committee is not charged with trying to rank the teams solely based on their abilities.”
Instead, the committee has to balance a number of competing factors, from avoiding conference matchups to managing prior games and maintaining overall bracket structure. Because of that, seeding reflects more than just team quality. That’s where analytics come in. Models like KenPom are designed to evaluate teams based purely on performance, without the constraints that shape the bracket, offering a clearer picture of how strong a team actually is.
“An analytics model is not constrained by that,” Pennington said. “They’re just ranking the teams based on what the analytics model tells them are the best teams.” That difference between seeding and analytics is where things start to get interesting, because when the two don’t align, it can reveal opportunities that aren’t obvious at first glance.
Where the bracket starts to break
One of the clearest examples this year is UConn. At the top of the bracket, most of the highest seeds closely match the top teams in KenPom, but UConn stands out as the exception. Despite earning a No. 2 seed, the Huskies rank 11th in KenPom, making them the only team on the top two seed lines whose placement doesn’t fully match its analytical profile.
That doesn’t mean UConn is guaranteed to lose early, but it does offer a data-backed reason to pause. And that’s often where better brackets begin, not by chasing certainty, but by questioning assumptions.
That same mindset becomes even more important when looking at individual matchups, especially the ones that tend to define the tournament. The 5-12 game has become synonymous with upsets, but even within that pattern, context matters. Not all 12 seeds are built the same, and not all favorites are equally stable.
Take Wisconsin and High Point. On paper, Wisconsin appears to be the stronger team, and its analytical profile supports that. But High Point's playing style requires nuance beyond the analytical profile. "They shoot a tremendous amount of threes," Pennington said. "That makes their games higher variance." In a single-elimination setting, that kind of volatility can be enough to flip a result, since one hot shooting performance can outweigh a season’s worth of consistency.
Why context still matters
Even the best analytics models have limitations because they cannot capture every real-world factor that affects a game. Injuries, for example, can completely change a team’s outlook. A ranking might reflect a full season’s performance, even if the roster looks very different by tournament time. Pennington pointed to North Carolina as a case where a key injury shifted the reality of the team, even though its metrics remained tied to earlier results.
Location can also play a meaningful role. Teams competing closer to home may benefit from crowd support and familiarity, subtle advantages that don’t always show up in the numbers but can influence outcomes. “This is why it’s not easy to pick winners of games, and why things like pure analytics models also aren’t going to guarantee you outcomes,” Pennington said. March Madness exists in that tension between what can be measured and what cannot, where data provides guidance but not certainty.
Building a smarter bracket
So what does it actually look like to use this approach? Pennington recommends starting with your instincts, then using analytics as a way to refine those decisions. Fill out your bracket, then compare your picks to models like KenPom and look for where your choices differ from the data. Those differences are often where the most important decisions happen.
If a pick goes against what the numbers suggest, it doesn’t mean it’s wrong, but it does mean it’s worth understanding why. That process helps turn guesswork into something more intentional. As Pennington puts it, “All models are wrong. Some of them are useful.” The goal isn’t to let the data make the decision for you, but to use it as a tool to sharpen your thinking and challenge your assumptions.
That mindset extends far beyond basketball. In fields like weather forecasting, models generate predictions, but human judgment is still needed to interpret them in context. The same balance applies when building a bracket.
The difference between being right and winning
There’s one final distinction that matters when filling out a bracket, and it’s where strategy comes into play. Picking the most games correctly is not the same as winning your bracket pool. If your goal is accuracy, minimizing bias is key. Choosing teams based on familiarity or preference might make the process more enjoyable, but it doesn’t necessarily improve your results.
If your goal is to win, the strategy shifts. “If you’re trying to win a bracket pool, you probably need to adopt a higher variance strategy,” Pennington said. In most pools, many participants will make similar picks, especially in later rounds, which means the outcome often comes down to a few key decisions. Selecting a slightly less popular team with a realistic chance to win can create a meaningful advantage and separate your bracket from the rest.
A better bracket starts with better thinking
March Madness will always be unpredictable, and that’s part of what makes it so compelling. But building a better bracket isn’t about eliminating uncertainty. It’s about understanding it. Looking beyond seeding, using analytics to challenge assumptions and paying attention to context can all help you make more informed decisions.
At Creighton, that kind of thinking is central to how students approach complex problems. Statistics isn’t just something applied during March Madness. It’s embedded across the academic experience. Every student completes coursework that builds statistical literacy, regardless of their field of study, from business to pre-health to the social sciences.
For students who want to go deeper, programs like Business Intelligence and Analytics and Data Science focus heavily on applying statistics in real-world contexts. In many courses, students are building predictive models, testing them and learning how to interpret results, the same types of skills used to analyze tournament outcomes.
Statistics may not guarantee a perfect bracket, but it can help you build a better one. And at Creighton, it’s one of the many ways students learn to turn data into smarter decisions.
