In this Freakonomics Radio episode, host Stephen J. Dubner examines the most scrutinized talent decision in American sport and asks why it still goes wrong so often. Stephen Smith, CEO of Kitman Labs, appears alongside a Hall of Fame quarterback, an NFL scouting director, and a sports psychologist, and makes the case for what changes when evaluation draws on years of connected data rather than a single moment of assessment.
Stephen describes a model trained on more than 15 years of NCAA performance and NFL combine data, and he is direct about what it can and cannot do. “We don’t have a crystal ball,” he tells Dubner. “The role is to reduce uncertainty.”
Why Listen
- Reduce uncertainty, don’t predict outcomes: why Stephen frames the job of a model as narrowing the range rather than calling the result.
- The traits that matter are the hardest to measure: processing speed, resilience, and coachability are proving stronger signals than size, speed, and arm strength.
- A broad set of inputs, not a single test: physical and physiological data, training history, match performance, and behavioral indicators considered together.
- “I think we need a connected pathway”: why the disconnect between the youth, collegiate, and professional games limits what any organization can see, and what a joined-up model would change.
Hear the full episode with Stephen J. Dubner right here, and explore more from Freakonomics Radio
