Sheldon Jacobson: March Madness and advanced analytics

March Madness begins March 15. Blue-blood programs like Villanova, Duke and Kansas are locks to make the tournament. Mid-majors like Loyola-Chicago and Murray State hope to pull upsets and advance to the second weekend. Virginia Commonwealth and BYU are just hoping for a spot in the Big Dance.

College basketball fans, and many others, participate in the tournament by filling out brackets. With 64 teams in the main bracket, after the First Four are played, there are 9,223,372,036,854,775,808 possible ways to do this.

Most people cannot fathom the magnitude of this number, over 9.2 quintillion. If every person in the United States filled in 900 brackets every second, it would take around one year to cover every bracket. Computations like this provide a glimpse into how advanced analytics can be used to provide insights into numbers that are beyond anyone’s imagination to grasp.

Not all these brackets have the same likelihood of occurring. Is there a way to parse through and identify the number of likely brackets? Advanced analytics can help here as well. Just like when we are at a grocery store and want to check out, we look for the lane with the shortest queue and is moving the fastest. That is advanced analytics thinking, though we rarely recognize it as such. That same thinking can be used to reduce the number of possible brackets down to the number of reasonably plausible brackets, though this number is still quite large.

Historical data since 1985 shows that picking teams that are highly seeded to reach the Final Four makes sense. What is the ideal combination of seeds? How about two No. 1 seeds, one No. 2 seed, and another team seeded between No. 3 and No. 12? Advanced analytics encourages the data to speak, and when building brackets, listening to the data makes sense. That is the same kind of data-informed thinking used to compute credit scores, which evaluate the likelihood that a loan will be repaid. A good credit rating means you are likely to be a good credit risk. Using historical data to identify trends can push the odds in your favor.

But what about upsets? They are what make the games fun to watch and draw the most attention. When a team seeded No. 12 defeats a team seeded No. 5, how can that be predicted? Upsets are guaranteed to occur. The hard part is knowing which teams they will be. This is about informed decision-making under uncertainty, an important component of advanced analytics, and a challenge we all face on a daily basis.

Take for example weather forecasting. It gives “chances” of rain and snow, yet these weather events may never occur. Nothing is for sure and absolute, yet we must make decisions to prepare for the unexpected. Looking for clues to help inform such decisions is like looking for team statistics that could predict an upset. Difficult? Yes. Possible? Certainly. Dealing and responding during the pandemic demonstrates how hard it can be to make decisions and choices with incomplete information.

Though advanced analytics may not be obviously seen directly on the basketball court during March Madness, every fan and spectator can use such thinking to fill out their brackets and perhaps enjoy the games.

Data analytics and decision-making under uncertainty are in the crosshairs of advanced analytics. Every upset is an opportunity to reveal the unexpected. Every team matchup is an opportunity to demonstrate advanced analytic insights.

Advanced analytics is ubiquitous across our lives, yet many rarely interpret it as such. With March Madness, advanced analytics is on full display for every person, young and old, to employ when filling out a bracket. When the tournament begins, advanced analytics may not explain why March Madness captures everyone’s attention, but it certainly is at work when we put pen to paper in our office pool.

Sheldon Jacobson is a professor of computer science at the University of Illinois at Urbana-Champaign and founder of the Bracketodds website, a STEM learning lab at the university.