Going Deep: AWS + NFL Next Gen Stats

Get an inside look at the team behind Next Gen Stats and discover how they use real-time player tracking data to generate advanced statistics and provide entirely new insights across the NFL. We break down the basics of their data pipeline, explore the difference between business logic and machine learning, and learn how AWS makes it all possible.

How does the NFL ingest and store billions of data points each season? The Next Gen Stats team walks us through their real-time pipeline—from stadium servers to Apache Kafka, which distributes events across hundreds of parallel processors. Discover why choosing the right database matters, and how the team uses services like MongoDB, Amazon S3, and PostgreSQL together to handle everything from raw tracking data to powering analytics dashboards with Amazon QuickSight.

The NFL Next Gen Stats team uses machine learning to create advanced statistics. In this episode, we'll show you how the team decides when to use ML versus rule-based approaches, how they use Amazon SageMaker to test and develop their models, and how they know when their models are ready to be deployed.

How do you predict the likelihood that a specific pass will result in a completion? And how does this stat enhance our understanding of the game? Listen in as the team discusses its collaboration with football analysts to translate domain expertise into machine learning features—from air yards to defender proximity. This episode also features a discussion of different ML model types (linear, tree-based, deep neural networks) and how the Next Gen Stats team settled on using XGBoost for Completion Probability.