The NFL announced the winners of the third annual 2020-21 Big Data Bowl, powered by Amazon Web Services, Inc. (AWS). The competition was hosted by NFL Network predictive analytics expert, Cynthia Frelund and aired on March 18 on NFL.com, along with the NFL's Twitch and YouTube channels.
Using Next Gen Stats powered by AWS, the 2020-21 Big Data Bowl called on both professional and aspiring data analysts to devise innovative and data-driven approaches to analyzing pass coverage in the NFL.
The winning group of Wei Peng, Marc Richards, Sam Walczak, and Jack Werner took home an additional $10,000 prize, bringing their competition total to $25,000. The group generated play outcome models for each frame of the data, as well as classified man versus zone coverage schemes to measure the before and after pass ability of each defender.
"The Big Data Bowl has changed how NFL clubs and their fans ingest and understand the game," said NFL Director of Football Data and Analytics Mike Lopez. "This year's event covered new ground in football analytics – defending the pass play. More than 250 participants submitted a unique approach, and the eight finalist teams represent the best of the best public football analytics work done to date. Today's event was the culmination of their hard work, as well as a celebration of how data is transforming what happens on the field and in front offices of the NFL."
Baltimore Ravens head coach John Harbaugh joined the program for a football analytics conversation alongside Frelund and Lopez. Harbaugh shared why the Ravens organization values data and analytics, advice for aspiring football data analysts, and his thoughts on the annual competition.
"This Big Data Bowl is a great way to broaden the perspective of the sport and to get more people involved in it in a fun way," said Harbaugh. "There are so many perspectives and ways to look at anything, especially football because it's a complex, crazy game."
Each finalist presented their respective algorithms to a panel of judges made up of Frelund, three former Big Data Bowl participants and a data scientist from AWS:
Last year's winning algorithm provided predictions for rushing play outcomes, and was adopted by the NFL's analytics team as one of this season's new Next Gen Stats. It was also used and shared with NFL clubs and media during the 2020 NFL season.
The competition also helps the league identify and develop future industry leaders. Dating back to the first Big Data Bowl in 2018, it has helped 15 participants secure jobs with either NFL clubs or affiliate vendors.
This year, the league added a mentorship program alongside the competition where 16 junior data scientists from diverse backgrounds were matched with experienced NFL analytics experts to curate a Big Data Bowl submission. Two participants in the mentorship program, Ella Summer and Jill Reiner, emerged as college finalists.
Below are other brief summaries of submissions finalists in the “Open” category presented during the Big Data Bowl:
Joe analyzed individual defender success while guarding a receiver making a cut or double move.
Dani Chu, Matthew Reyers, Meyappan Subbaiah, Lucas Wu
The group of Meyappan, Dani, Matthew and Lucas isolated various parts of defensive coverage throughout a pass play to evaluate defender contribution.
Matthew Gartenhaus, James Venzor
The group of James and Matthew modeled six different defender attributes which take place during different stages of a pass play.
Tony ElHabr, Marschall Furman, Sydney Robinson, Asmae Toumi
The group of Asmae, Marschall, Sydney and Tony created target and completion probability models to build a defender effectiveness metric.
Below are brief summaries of submissions finalists in the “College” category presented during the Big Data Bowl:
Zach Bradlow, Zach Drapkin, Ryan Gross, Sarah Hu – University of Pennsylvania
The group of Zach, Zach, Ryan and Sarah created a relative skill rating system based upon success in coverage matchups to measure defender skill.
Ella Summer – University of Virginia
Ella used statistical modeling to isolate individual defender effects on target and completion probability.
Jill Reiner – Denison University
Jill used statistical modeling to determine which defenders are best at averting targets, closing on targeted receivers and defending passes, and clustered each based on their strengths and weaknesses.
To watch this year's presentations click here. To learn more about the Big Data Bowl, visit https://operations.nfl.com/gameday/analytics/big-data-bowl/.