ipitytheblue
...
The percentage accuracy numbers are also misleading. 80% or so of ball/strike calls are not on/near the edges of the zone and are easy to make. So he got 5 or 8 out of around 24 pitches incorrect. Missing between 20-35% of borderline calls in a world series game shouldn't be praised by anyone. AI (once programmed correctly) would miss zero.
The scorecard should target a zone of close calls, i.e. where the ump's call actually "matters". Everyone in the stadium could call a middle-of-the-zone strike or a ball in the dirt, not to mention all the swinging strikes and foul balls etc. Half the reason the ump is standing there is for the calls within a few inches of the edges and corners. I'd love to see something approximating a Kappa statistic on the scorecard, i.e. the % of correct calls beyond 50% (random) that the ump made. That way, 100% would be perfect, 0% would be me on my couch, and some number in between would be the tangible effectiveness of the ump. Much more meaningful than total % correct.