Tracking the ball

The problem

Put an inertial sensor inside a football and you get a firehose of acceleration and rotation. Turning that into spin rate, velocity and throw quality in real time is a signal-processing problem.

My role

Algorithm engineer.

Approach & outcome

• Algorithms and data pipelines for the Wilson X line of connected footballs and basketballs.
• Field studies with NFL and NCAA teams on whether ball-level data could give athletes and coaches next-generation analytics.

What the field taught me

The lab is clean; practice isn’t. Balls changed hands constantly, so just knowing who was throwing, and when, was a problem of its own. The Bluetooth radio inside the ball wasn’t always strong enough to capture a throw cleanly. And the sensors themselves shifted the ball’s mass and moment of inertia just enough that professionals could feel the difference in their hands.

That last one stuck with me. The best algorithm in the world doesn’t matter if the instrument changes the thing you’re trying to measure. Now I start every sensing problem by asking what the measurement costs the person being measured.