Amazon Web Services (AWS) has added six insights to the NFL’s machine-learning-based Next Gen Stats platform. The stats will appear in live broadcasts and online throughout the 2020 season, offering fans a better understanding of the game, their favourite players and teams’ performances.

Covering everything from the expected yards a team will gain on a rush, to the probability a team will win a game, the stats combine on-field action and cloud technology, including machine learning, to engage fans in real time from season kickoff to the Super Bowl.

The NFL has used AWS compute technology for its Next Gen Stats platform since 2015, and began using AWS machine-learning services to enhance the stats in 2017. These stats help teams, players, broadcasters and fans better understand and appreciate football, using more than 300 million data points collected each season to generate content, including data from sensors in players’ gear and game balls, as well as reference points throughout NFL stadiums.

The stats powered by AWS for the 2020 season are:

• Expected rush yards – Helps showcase the speed, elusiveness and abilities of running backs by predicting how many yards a rusher is likely to gain with the ball upon handoff. A team of independent data scientists created the algorithm that powers the stat during the 2020 Big Data Bowl, the NFL’s annual sports analytics contest.
• Route classification – Automatically identifies in real time the route a player runs on a pass play – go, post or out – providing the groundwork for a new understanding of offensive strategy and tendencies, while also delivering metrics to rank the performance of individual players, team offences and team defences by route.
• Expected points – Calculates the likelihood of either team scoring next and by what means (touchdown, field goal and safety), enabling analysts, coaches and fans to quantify the impact of an individual play on helping a team score.
• Win probability –Assigns a win likelihood percentage to each team, adjusted in real time throughout the game, to identify changes in momentum and the plays that are most likely to impact game outcome.
• Expected yards after catch –Predicts how far a receiver is likely to gain with the ball upon completing a catch. Similar to expected rush yards, this helps identify how receivers can improvise and elude defenders to gain yards after receiving the ball.
• Field goal probability –Quantifies the likelihood of a team scoring a field goal given the distance of the kick and weather conditions.

The NFL will rely on AWS services to generate these insights and scale its analysis of action on the field, processing more than one million data points during every game. Leveraging Amazon SageMaker, AWS’s managed service for building, training and deploying machine-learning models quickly, the NFL can rapidly iterate and train multiple variations of the machine-learning models that power the stats to achieve the best performance.

In addition, the NFL uses Amazon Simple Storage Service (Amazon S3) to store on-field data for quick and secure reference, and Amazon Elastic Compute Cloud (Amazon EC2) to process those data, making it possible to provide real-time insights to fans at home. The full set of Next Gen Stats from 2020 and earlier NFL seasons are available for use in live broadcasts and afterwards in online replays.

“This season’s Next Gen Stats powered by AWS represent the NFL’s continued investment in cloud technology to improve not only the fan experience for millions of people around the world, but also the way the game is played,” said Matt Swensson, vice president, at the NFL. “Our partnership with AWS and the use of machine learning have enabled our fans to better understand the remarkable performance of the athletes on the field. They have also uncovered opportunities for NFL teams to use these data to improve their performance by breaking down the action in new ways and analysing plays at a deeper level after the game.”

Greg Pearson, vice president at AWS, added: “AWS is proud to continue its partnership with the NFL to bring engaging data to life through Next Gen Stats and enhance the football experience for everyone in the NFL community – fans, players, coaches, broadcasters and more. In the 2020 season, we’re going to expand on the already compelling set of Next Gen Stats to underscore the talent on the field and help people better understand and appreciate the strategic decisions needed to be made and the incredible athletic feats on the field.”