IRMAS Student Uses Stat Modeling to Predict NHL Playoffs
By Reginald Allison II
“Defense wins championships” is one of the oldest clichés in sports. In fact, legendary football coach Paul “Bear” Bryant is the person credited with the quote. Sports media has since vehemently debated and widely perpetuated the adage, turning it into gospel in some sports. But is it true wisdom or a mere platitude? Nicholas Laffey, a senior majoring in math and economics, has set out to put an end to the debate.
As part of the University of Alabama’s Catherine J. Randall Research Scholars Program, Laffey is working with assistant professor Dr. Brendan Ames to analyze and predict the performance of hockey teams. “Currently I am developing various models to predict playoff success of NHL teams. These models use a machine learning technique called sparse regression that has a few advantages over ordinary least-squares regression,” Laffey said. “Being able to better understand what factors are responsible for a team succeeding in the regular season versus the playoffs can greatly influence the way in which teams are built, players are drafted, and how salary cap space is allocated.”
The Randall Research Scholars Program, formerly Computer-Based Honors, is a nationally recognized undergraduate research program, which pairs exceptional students directly with leading research professors and cutting-edge computing technology to conduct research in a variety of disciplines. It is a component of UA’s Honors College. At its 1968 launch, the program was the nation’s first university-wide undergraduate research program. Over the past 48 years, more than 1,000 students have participated in the program.
Laffey first learned about the Randall Research Scholars Program when he applied to the university and thought the opportunity for undergraduate students to work one-on-one with professors on research was a fantastic opportunity. “I have always had a strong interest in utilizing computer technology to solve various problems in math, economics and finance, so I thought the opportunity to learn all of these skills was invaluable,” he said.
He began working on this particular research topic during his senior year as a means to connect the theories he was learning in class to their real-world applications. “Normally, a lot of the math you learn in class can seem too abstract and unconnected to real-world problems. So, I thought it was a really cool thing to be able to apply math and statistics to something so many people are passionate about like sports,” Laffey said.
However, working on this research has not been without its challenges. “I created the models in this project with R, a programming language I did not know when I first started, and had no clue where to begin with understanding what sparse regression meant,” he said. “Working with Dr. Ames, I’ve learned how to approach a subject that I know very little about. I was able to read the relevant literature and piece together a plan for executing this project. I believe this experience will be a great asset in tackling new problems in my career going forward.”
So, what do the numbers say? The results suggest that regular season defensive performance is significantly more influential in predicting playoff performance than in the regular season. In addition, statistics reflecting a combination of both offensive and defensive ability exhibit increased influence in models fit to playoff performance. Laffey said that one interpretation of the results is that teams need to have good offenses to make the playoffs, but once they are in the playoffs, defense is the distinguishing factor. These findings lend credence to the cliché that defense does, in fact, win championships.