The NFL and Machine Learning: A Touchdown for Technology?

In an increasingly competitive industry, professional sports teams are seeking innovative ways to create unparalleled fan experiences. The NFL is now leveraging big data to optimize a nation's tradition.

“Perfection is not attainable. But if we chase perfection, we can catch excellence,” pronounced National Football League (NFL) legend Vince Lombardi. Each game day, NFL athletes play through sweat and tears to give millions of fans unparalleled game experiences. However, NFL ratings and fan satisfaction are at an all-time low.[1] Moreover, fans now view games through new mediums. In 2017, viewers watched more than 10 billion minutes of video across the NFL’s digital and social platforms.[2] At a crossroads, the NFL must “chase perfection” to satisfy its evolving customer base and return to ratings growth.

Data scientists have recognized the predictive value of machine learning in the sport.[3] For example, machine learning models have been able to predict football turnovers with surprising accuracy.[4] The NFL has noticed the value in these methods as well. In 2015, the NFL began Next Gen Stats (NGS) to revolutionize the sport with machine learning.[5] The visual below shows how the NFL’s NGS technology was built over time.

The NFL’s Next Gen Stats technology wasn’t build overnight. Source:

Every week of NFL football generates 3TB, equivalent to 1500 hours, of data.[6] In the short term, the NFL wants to turn this data into value for teams and fans.[7] The NFL has embraced radio-frequency identification (RFID) tags as one strategy for achieving this goal.[8] RFID-implanted equipment captures live data, and NGS puts this data to work.[9] NGS creates adaptive models based on historical data (e.g. past routes run, field data, weather conditions, etc.) that provide key decision-makers with useful in-game strategies.[10] Once implemented, this machine-learning technology will instantly analyze a play’s formation, route, and key identifiers in real time.[11] For example, machine learning can help a coach determine if a quarterback made a good decision on a pass right after the play. Real time analysis replaces “Monday morning quarterbacking,” said NFL Senior VP and CIO Michelle McKenna.[12] After the game, coaches also receive insights that include fitness summaries, heat maps of player locations, and relative speed and distance play diagrams for each player to help them adjust future game strategies.[13]

Beyond the game itself, the NFL leverages data to create value for the increasing number of digitally-savvy fans. Each week, the NFL publishes the “Next Gen Stats: Hidden numbers” that presents data-informed predictions for teams and individual players across a variety of performance metrics.[14]

Plans also include innovating the fan experience with machine learning. The NFL has developed educational tools for commentators and television broadcasters to better use data to drive more rewarding programming for viewers.[15] Matt Swensson, NFL VP of Emerging Products and Technology said, “Machine Learning and other computations that could take months to refine now take weeks or days, allowing us to engage, inform and excite fans in new and unique ways.”[16]

However, achieving the promise of machine learning poses challenges. To add value with data, the NFL needs to move swiftly to assess skills gaps, invest in training, and make smart hires. Coaching staff education must rank high on the priority list. In order for data deployment to achieve success, coaches need to buy into the idea. Right now, many coaches remain skeptical. “All that stuff is good to have. But it’s on film, too, and the film don’t lie,” said former Indianapolis Colts Coach Chuck Pagano.[17] Others echo Pagano’s concerns, including Seattle Seahawks Coach Pete Carroll, “We don’t have enough background yet to kind of make sense of it, how it’s helping us at all.”[18] Earlier this year, NFL union executive Ahmad Nassar estimated that “30 of the 32 franchises ignored the data altogether.”[19] Data can help coaches win games, but coaches need to believe that first to make it happen. As of March 2018, NFL teams have access to data of the opponents. [20] As a result, coaches will be incentivized to understand how this data can be used in game strategy as a source of competitive advantage.

The NFL has only recently adopted machine-learning technologies, and there are many unanswered questions. For example, the use of data may overwhelm and confuse fans. What are the most important NFL data analyses from machine learning that can not only interest, but increase the loyalty of viewers? Assessing an additional key stakeholder, how can the NFL accelerate adoption of NGS metrics among coaches? As the NFL progresses their understanding of the role machine learning can play in the organization, are there comparable organizations to take lessons from? In this early stage of machine learning integration, the NFL has every opportunity to thoughtfully engage machine learning in the most value-adding way for all. (word count: 749)

[1] Joe Harpaz, 3 Ways Artificial Intelligence Can Save the National Football League, Forbes (Jan. 11, 2018),

[2] NFL Digital Media Unveils New Product Features and Content Offering Fans a Deeper Connection to the Game, NFL Communications,

[3] K. Pelechrinis & E. Papalexakis, Footballonomics: The Anatomy of American Football: Evidence from 7 Years of NFL Game Data, 11(12) PLoS (2016).

[4] J.R. Bock, Empirical Prediction of Turnovers in NFL Football, 5(1) MDPI (2016).

[5] NFL Next Gen Stats, NFL Ops,

[6] Jason Hiner, How the NFL and Amazon unleased ‘Next Gen Stats’ to grok football games, Tech Republic (Feb. 2, 2018),

[7] Next Gen Stats: Powered by AWS, Amazon Web Services,

[8] See J. Pletz, A Victory Lap for Zebra? 41(38) Crain’s Chicago Business (2018).

[9] Taylor Soper, Microsoft Will Show ‘Next-Gen Stats’ on NFL App Thanks to RFID Chips Worn by Players, GeekWire (Aug. 7, 2015),

[10] Michelle-Doyle McKenna Shares How the NFL Can Take ‘Next Gen Stats’ to the Next Level, YouTube (Nov. 30, 2017),

[11] Id.

[12] Id.

[13] Id.

[14] Nick Shook, Next Gen Stats: Hidden Numbers That Could Define Week 9, NFL (Nov. 1, 2018),

[15] Michelle-Doyle McKenna Shares How the NFL Can Take ‘Next Gen Stats’ to the Next Level, YouTube (Nov. 30, 2017),; see also D. Kudenko & M. Zheng Automated Event Recognition for Football Commentary Generation, 2(4) International Journal of Gaming and Computer-Mediated Simulations (IJGCMS) 67-84 (2010); see also Paolo Del Nibletto, NFL Adopting Machine Learning, (Dec. 18, 2017),

[16] Next Gen Stats: Powered by AWS, AWS,

[17] Kevin Seifert, NFL Coaches Skeptical on Benefits of Chip-Generated Game-Day Data, ESPN (Jul. 24, 2017),

[18] Id.

[19] Joe Lemire, NFL To Distribute Full League-Wide Zebra Tracking Data, ESPN (Mar. 5, 2018),

[20] NFL Next Gen Stats, NFL Ops,


IoT and Me: Nest, Machine Learning, and the Smart Device Revolution


SEPHORA: Developments in Machine Learning and Augmented Reality towards customer personalization.

Student comments on The NFL and Machine Learning: A Touchdown for Technology?

  1. Thank you for this great essay on how NFL is leveraging big data and machine learning. For the fans, machine learning could be taken to another level where game stats data can be customised for each fan as different groups of fans may be more interested in one aspect of the game compared to another. I agree that for more adoption of big data and machine learning in NFL, the coaches will need to buy in and use the information to further improve their team’s performance. Perhaps a comparable sport could be cycling, where the cycling teams are using the data gathered from their cyclists to improve either equipment design or team strategy. In addition, cycling has also leveraged the digital age to provide real-time statistics to further engage their fans:

  2. This is a great topic – I was excited to read as soon as I saw it. I appreciate your thorough and informative portrayal of what the NFL is doing with respect to data collection and machine learning.

    I think an additional area to explore here is what the collection of such granular data will mean for sports betting, particularly given this is now legal in several states. The availability of data like speed at which the quarterback throws opens numerous in-game betting options – which can be seen as a great way to further engage fans, or a risk that players will be tempted to alter the game in ways that are extremely subtle.

    I enjoyed and agreed with your thoughts on teams’ and coaches’ use of data as well. To add to those, I wonder if the increased data availability and sophistication of analysis will make teams more likely to take actions that are unusual now, but widely acknowledged from an analytics perspective as having value (e.g., going for it on fourth down, which coaches “should” be doing more). I’m curious to see how having access to other teams’ data affects interest levels in machine learning and analysis – all the sudden if one team is using the data effectively, everyone will need to also to catch up. Lastly, I’m fascinated by the thought of machine learning picking up on plays and coverage schemes real-time.

  3. Great topic, can I use it to help my struggling fantasy team? This could be the beginning of the NFL’s “Moneyball” revolution. I wouldn’t be surprised to see the coaches mentioned above who don’t embrace the data replaced with quantitative coaches that do. NFL teams could start scouting for motivational data scientists to run the offense. Teams who embrace will benefit in the short term and the slower teams will eventually catch up and there will be a new equilibrium.

    In regards to fan loyalty, I go back to my first question of how can this help fantasy football. It would be great if the NFL made its data open source so fans could find their own new and creative ways to use it. Would also love to see it more mixed in and integrated with the Madden gaming community which is also a growing industry.

Leave a comment