Machine Learning on Thursday Night Football
The 6th Annual Awards
- Best Use of Technology
ABOUT THIS ENTRY
Prime Video’s Defensive Alert is the first machine-learning model to be used in live TV analysis and is exclusive to Thursday Night Football. Before each snap, the AI-powered model helps fans better understand a Quarterback’s process by identifying defenders who are likely to blitz… live. Once the model recognizes a unique pass-rush threat, a red orb graphic overlay appears around him to identify him as a Defensive Alert. This happens in real-time and is dynamic to account for player movement. This feature is unlike any previous broadcast tool as it provides fans additional analysis independent of the announce team as producers do not manipulate, edit, or alter the overlays once the AI model is activated. The model was trained on a historical dataset of over 50,000 NFL plays and combined with human expert knowledge from former NFL players and coaches.
How does this represent “Excellence in Engagement”?
Defensive Alerts marks an industry achievement as it is the first machine learning tool to be used in a sports broadcast to help augment the role of an analyst. Its successful launch proved that groundbreaking technologies, such as AI, can be additive to a sports broadcast (and also fun!). Defensive Alerts also proved that even the most advanced technologies need to create an end product that is intuitive and simple to understand. In this case, almost all viewers immediately recognized that a red orb encircling a defender meant that he posed a threat to the offense. This feature also reinforced the idea that technology cannot be created just for the sake of technology but rather to elevate the fan experience. A viewer does not need to understand football concepts, the inner workings of a machine learning model, or even need the announcers to guide them through this viewing experience. Defensive Alerts is not only ground-breaking but creates a road map for other broadcasters to follow as they continue to iterate the sports watching experience.
Objective
Thursday Night Football and its weekly Alternate broadcast “Prime Vision with Next Gen Stats” aims to make sports fans smarter. This tenet is the inspiration behind Defensive Alerts. In the NFL, there is a game-within-a-game before each snap as teams try to align in advantageous positions. This is one of the most challenging parts of the game to understand, even to players. For years, analysts have been trying to help fans better understand this aspect of the game, but lacked real-time visuals. The goal of Defensive Alerts was to create a real-time analysis tool to address this need. Initially, this feature was built on a set of logic that most football teams follow. However, after analysis of different rule sets, it became obvious that there was an opportunity to go even further and allow a machine-learning model to reveal what a human cannot.
Strategy & Execution
Defensive Alert was created to help sports fans better understand the game of football by providing deeper insights into the game they love. It started with the question: “What if we could take viewers inside the mind of the quarterback?” This AI-powered model emulates the pre-snap cadence of a quarterback and his offensive line by identifying coverage defender(s) that are likely to rush the QB on passing downs. Andrew Whitworth, 16-year NFL veteran, describes each NFL play as a puzzle and that before each snap, the quarterback and offensive line are trying to solve that puzzle. They rely on years of experience and thousands of hours of film study to find little clues that help solve each problem. Defensive Alert allows viewers to skip all the preparation and understand the puzzle’s solution as well as, or even better, then the players on the field.
Celebrate in NYC
Join us for the 6th Annual Hashtag Sports Awards™ on the evening of June 6, 2024 as we celebrate finalists and reveal the winners of this year's awards live during the show.
