How Machine Learning and Analytics Are Transforming the NFL
This article originally appeared on WSJ. Custom Studios
In 2018, the most popular TV program wasn’t a critically acclaimed drama or a hot new reality show. It was football.
NFL games accounted for 46 of last year’s top 50 telecasts, averaging 15.8 million viewers over the season. For fans, the sport’s continued appeal may have a lot to do with the chess-like combination of strategy, preparation and instinct that goes into every play. No factor can be overlooked for its potential impact on how those plays pan out: everything from changes in team rosters to the conditions on the field (i.e., indoor vs. outdoor, day vs. night, not to mention wind and precipitation).
While the NFL has tracked a wide variety of statistics since its inception, for decades these metrics were relatively rudimentary—like simply counting statistics that didn’t necessarily reveal the full scope of what happened during plays. Recently, the NFL realized it needed a more advanced system to collect data and make sense of it. Such a system could reveal insights about game dynamics for both fans and players—for example, the results of particular player lineups or the factors that impacted a player’s performance. The end goal: to create customer loyalty for the NFL and fuel diehard fans in their quest to better understand the game.
Today, the NFL’s Next Gen Stats (NGS) program uses sophisticated tracking technology collected via RFID devices in the shoulder pads of every player and embedded at each of its stadiums. These devices capture data about which players are on the field at a given moment, their location within inches, and the speed and direction in which they move. This treasure trove of data represents a tremendous resource for the league’s 32 teams, multiple media partners and approximately 180 million fans worldwide.
Partnering with Amazon Web Services (AWS), the NFL is leveraging the power of its data through sophisticated analytics and machine learning. “Machine learning is unlocking potential for us to do more than we otherwise could, in a timely manner with a high degree of confidence,” says Matt Swensson, vice president of emerging products and technology for the NFL. “We had a lot of stats and wanted to find the best way to leverage them. We’re taking in so much data now with the tracking system that we’re able to use machine learning to understand what elements are relevant and what are not.”
Powered by the machine learning tool Amazon SageMaker, the NGS platform allows the NFL to quickly and easily create and deploy machine learning models capable of interpreting the gameplay. One example is NGS’s Completion Probability metric, which integrates more than 10 in-play measurements ranging from the length and velocity of a specific pass to the distance between the receiver and the closest defenders—as well as the quarterback and nearest pass rushers.
Using Amazon SageMaker to easily build, train and run these predictive models helped reduce the time to get to results from as much as 12 hours to 30 minutes. And as Swensson points out, with SageMaker, the NFL doesn’t need to arm itself with teams of data scientists—its engineers can get up and running quickly. “We don’t have to reinvent the wheel every time we want to do something,” Swensson says.
The results help fans understand why some passing plays are more difficult than others and provide more meaningful understanding of the game itself. These insights can quickly be used by the NFL and its media partners to enhance broadcasts and online content, or even to educate and excite fans inside the stadium. “I’ve gotten a lot of positive feedback from fans saying, ‘Wow, how did they complete that pass?’ We’ve been able to quantify it and compare it to other passes, and that’s been a real value-add for fans because it creates context for what’s happening in the game,” Swensson says.
Of course, data is only useful when it can be quickly and easily accessed. Using the business intelligence tool Amazon QuickSight, the NFL is able to gain greater insight internally while also opening a window for fans to engage with data. “It allows us to run extremely fast queries to ask questions and surface the answers on dashboards,” Swensson says. “We provide dashboards to our clubs, to our broadcasters, and to our editorial folks and fantasy football writers at NFL.com.”
Those dashboards, which used to take hours or days to build, can now be created in minutes and can also include any number of relevant filters. “It’s allowed us to avoid writing a lot of code every time we want to show information,” Swensson says. “It’s much more efficient.”
Additionally, the NFL can then take these insights and apply them to different parts of the organization, helping coaches create better game plans and even finding ways to improve player safety. “The more information you have, the better you can identify patterns in play,” Swensson notes. Those patterns, identified through machine learning, could be the keys to better understanding where players are more likely to get injured and to help design rules to mitigate risk.
The end result is a better experience for fans, players and teams—all in real time. It’s nothing less than the next generation of NFL football, and it’s powered by next-gen analytics and machine learning.