The National Football League (NFL), akin to numerous professional sports industries, is embracing the potential of artificial intelligence. Through a strategic partnership with Amazon Web Services (AWS) known as “Next Gen Stats,” the NFL is embarking on a journey to harness intelligent algorithms and cutting-edge data collection tools to extract profound insights from games and unravel intricate player performance patterns. AWS was inspired by submissions to the 2023 Big Data Bowl, an annual software competition organized by the NFL, prompting the creation of a new category of analytics focused on “pressure” in football.
With AWS at the helm, AI-powered algorithms have been developed to analyze player behavior on the field, discerning the nuances of a defender’s aggression, speed, and a quarterback’s responsiveness. This granular data allows for the quantification of pressure, offering game analysts a deeper understanding of the strategic elements influencing plays. These advanced analytics transcend traditional statistics, providing a richer context. While conventional data might reveal whether a rusher reached a quarterback, it may fall short in explaining the intensity of the encounter. This is where the concept of “pressure probability,” tracked by “Next Gen Stats,” dives into the finer details.
The collaborative efforts of AWS and the NFL have culminated in machine-learning models that furnish data in three pivotal areas of gameplay. Firstly, AI has been trained to identify blockers and pass rushers in passing plays. Secondly, the tool has been equipped with the ability to quantify “pressure” during a game. Lastly, a process has been developed to detect individual blocker-rusher matchups. This AI-tracking technology empowers professionals in the football league with invaluable player statistics that can aid scouts and coaches in selecting new players. For instance, understanding which player effectively blocked or resisted a rusher can inform decisions about their suitability for an offensive lineup.
In the world of football, evaluating the performance of offensive players and the rushers who tackle them can prove challenging, even for seasoned experts. Player reactions unfold in fractions of a second, and their performance in these lightning-fast exchanges is difficult to track, let alone quantify. Metrics such as how close a defender came to the offensive lineup can offer coaches insights into the strength of their plays.
The NFL collects data for these AI-driven processing tools using proprietary field-installed equipment. In each NFL venue, an array of 20-30 ultra-wideband receivers are embedded in the field, while radio-frequency identification (RFID) tags are affixed to players’ shoulder pads and other game equipment, including balls and goalposts. These data transmitters capture information relayed in real-time through a graphic neural network model (GNN). AI then transforms this data into meaningful insights.
These insights manifest as interactive graphics on the Next Gen Stat game landing page. Users can access detailed breakdowns of individual player movements in any given game using 2D models and graphs. For example, they can track the trajectories of players and the ball during a 40-yard passing play in a game between the San Francisco 49ers and the New York Giants on September 21.
While the AI tool resides on AWS infrastructure, the final product is the result of a multifaceted partnership involving the NFL, Zebra Technologies, and Wilson Sporting Goods. The Next Gen Stats initiative, initiated in 2017, now comprises a data pipeline housing historical data from every pass play since 2018.
In parallel, AWS engineers are diligently working on automating the identification of blockers and rushers, aiming for AI models to autonomously recognize players’ roles on the field. Presently, this information is manually collected, a process prone to labeling errors and time-consuming for humans to generate.
