St. Olaf College | News

Three Oles are big winners in NFL’s Big Data Bowl

Marc Richards '16 (left) and Sam Walczak '14 (right), pictured here with Minnesota Timberwolves mascot Crunch the Wolf, are two members of the team that won the National Football League's Big Data Bowl.
Marc Richards ’16 (left) and Sam Walczak ’14 (right), pictured here with Minnesota Timberwolves mascot Crunch the Wolf, are two members of the team that won the National Football League’s Big Data Bowl.

When it comes to analyzing big data from the country’s biggest sports league, three St. Olaf College alumni are big winners.

Marc Richards ’16, Sam Walczak ’14, and Jack Werner ’16, along with teammate Wei Peng, are the champions of the National Football League’s Big Data Bowl

The annual contest challenges talented members of the analytics community to explore statistical innovations in football. The NFL captures real-time data for every player, on every play, in every situation — anywhere on the field — and contestants in the Big Data Bowl use traditional and Next Gen Stats to analyze and rethink trends and player performance.

Jack Werner ’16 (left) and Sam Walczak ’14 are two members of the team that won this year’s NFL Big Data Bowl.

Richards, Walczak, and Werner had bonded at St. Olaf over an interest in sports analytics, and they started Model 284, a website to showcase their various sports-based data science projects, after they graduated. Together with Peng — whom Richards met in their Ph.D. statistics program at the University of Pittsburgh — they decided to enter this year’s Big Data Bowl, which challenged participants to use the NFL’s player tracking data to develop better methods for analyzing defensive performance in pass coverage. 

As a Star Tribune story about their success notes, “Peng, Richards, Walczak, and Werner first built a model to determine if a player was in man or zone coverage, based on how they were lined up at the snap and how they moved once the ball was snapped. That allowed them to group coverage assignments into different clusters, build a model predicting how many passes should be completed based on how close a defender is in coverage, and evaluate player performance against those expectations.”

Their winning analysis netted them a total of $25,000 in Big Data Bowl prize money.

The team will meet with current students in the St. Olaf Sports Analytics Club this month to discuss their Big Data Bowl work. Ahead of that, they took a few minutes to share how they explored their interest in sports analytics during their time at St. Olaf, why they decided to enter this NFL contest, and the most surprising data point they took away from their analysis (spoiler: it has to do with the Minnesota Vikings!).

What prompted you to enter this competition?
Walczak: We are all big NFL fans, and we’d seen the previous Big Data Bowls put on by [NFL Analytics Director] Michael Lopez and the NFL. We have worked on a number of sports analytics projects together in the past, so tackling this type of problem was a natural fit for us. We all have been busy with life and haven’t really had the time to work on a lot of sports analytics projects in the last year or two, so we thought this would be a great opportunity to get working again and stay in touch during the pandemic.

How did you decide what approach to take in analyzing the data?
Walczak: The prompt for the Big Data Bowl this year was very open-ended, and more or less directed us to evaluate defensive coverage (i.e., how can we better understand and evaluate a defense’s ability to cover receivers?). The NFL supplied their new player tracking data for this contest, which we had never worked with before. We thought first about the issues with the traditional metrics for this question, which don’t leverage player tracking data. For example, only one defensive player is typically evaluated on each play (the defender closest to the receiver when the pass arrives). To overcome these issues, we created our own framework for evaluating individual defensive coverage performance. It’s a three-step process for each defender: (1) find out whether he is in man or zone coverage, (2) identify which offensive player(s) he is responsible for covering throughout the play, and (3) apply a number of metrics we developed to evaluate how well he covered his receiver.

What was the most interesting thing you took away from this project?
Walczak: The data for the Big Data Bowl was for the 2018 NFL season. Being big Vikings fans, we were curious to see how some of the Vikings players performed. Xavier Rhodes, whose performance on the surface (and based on other analysis) seemed to drop off significantly in 2018, actually ranked in the 75th percentile in man-to-man coverage ability using our metrics. That was much better than we would’ve expected.

Were you surprised to win the Big Data Bowl?
Walczak: Incredibly surprised! We did feel that our work was innovative and thoroughly explained, but there were so many awesome submissions ranging from students to professors to industry data scientists from Microsoft. It was really an honor to be a finalist, let alone the grand prize winners.

The three of you were interested in sports analytics even as St. Olaf students. How did you explore that interest as an Ole?
Walczak: In my senior year at St. Olaf, I did an independent statistics research course with Professor [of Mathematics, Statistics, and Computer Science] Matt Richey where we built a model to predict the NCAA Men’s Basketball Tournament (March Madness). That spring (2014) we applied the model to a “live” NCAA tournament for the first time. I remember sitting in Regents 284 (our favorite room, and the inspiration for our website name “Model 284”) the day the bracket came out, anxiously awaiting the model’s predictions, and it displaying UConn as its predicted champion. UConn, a #7 seed, was a huge underdog that no one was talking about, but ended up going on a Cinderella run to win the whole tournament. Suffice it to say, I was all-in on this whole sports analytics thing from that moment on. We actually still keep up this model and post its predictions  on our website

Richards: Being a college hockey player and math major at St. Olaf, sports analytics naturally interested me. I then met Sam through mutual friends and we immediately bonded over a passion for sports and statistics. Later, Jack and I took a number of courses in math, statistics, and computer science and were both in the CIR (Center for Interdisciplinary Research). Jack also shared our similar passion for statistics and sports, and we worked on a number of projects in our statistics courses in sports analytics — for example, building a model that projected college basketball performance to the NBA.

Werner: I’ve always been interested in math in sports. I was lucky to attend St. Olaf, where the Mathematics, Statistics, and Computer Science Department provides tons of opportunities to apply stats and data science to real projects. I was involved in the CIR and Math Practicum, which are both great examples of that. I incorporated sports whenever I could, like working on an NBA-related class project with Marc and doing a baseball research independent study with Professor Richey.

What do each of you do now for a career/graduate school?
Walczak: I work as a reinsurance broker and catastrophe modeler at Holborn here in the Twin Cities. I am also in the Data Science Master’s Degree Program at the University of Minnesota. 

Richards: I am currently a Ph.D. student in the Department of Statistics at the University of Pittsburgh. I work under Dr. Lucas Mentch, and we are studying applications of spatial statistics in crime across U.S. cities as well as developing methods for classical statistical inference in machine learning problems. In between my Ph.D. studies and undergraduate degree, I worked for three or four years as a catastrophe risk analyst modeling natural disasters for a reinsurance brokerage while working on sports analytics at night with Sam and Jack.

Werner: I am currently a data scientist at Wells Fargo, where I use text analysis and natural language processing techniques to identify complaints in customer feedback.

How did St. Olaf help shape your interests and path?
Walczak: The statistics and math courses I was able to take as an undergraduate at St. Olaf were instrumental in developing my analytical and problem-solving skills. The CIR was also an incredible opportunity to work with a team to solve a complicated problem. I was lucky to cross paths with so many awesome people at St. Olaf, which shaped who I am today in many ways. I wouldn’t trade it for anything. 

Richards: I can’t speak enough about my experience at St. Olaf. I got the opportunity to learn and work with some of the best teachers and researchers that I have encountered in my professional and academic career in Dr. Paul Roback, Dr. Mary Walczak, Dr. Katie Ziegler-Graham, and many others. Beyond academics, I really enjoyed being a part of the St. Olaf Men’s Hockey program. Being a student-athlete at a first-class institution has prepared me extremely well for a successful academic and professional career.

Werner: I loved my time at St. Olaf! I liked being able to take a wide variety of classes, liberal arts-style. And once I settled into my math and stats path, I found those departments to be phenomenal. The professors were all very helpful and interested in providing opportunities to students. By the time I was done, I had gotten experience as a stats consultant, presented at a conference, co-authored a paper, done an independent study, and spent a lot of quality time coding R in the CIR room. It set me up well for the jobs I’ve had post-Olaf (and for a fun, prize-winning hobby!).

For more analysis of sports through math and models, follow these Oles on Model 284.