This is what 8 million data points looks like in production
8.3 million pitches. Every single one tagged with velocity, location, and outcome. That's not a dataset. That's a living system.
Umpire profiles rebuilt every night at 10:30 PM MT. Yesterday's model was already less complete than today's.
You're not tagging pitches. You're tagging context. Batter, count, game state, location. Each pitch earns its place in a profile.
95K games. 2015 to today. The longitudinal depth alone changes what questions you can ask.
Most ops teams maintain a system like this. You're still building it — and building it faster than they ever did.
Book the $750 AI ops audit →