Entering the final week of the first year of the ABS Challenge system in Major League Baseball, appeals had overturned 5,483 ball-strike calls. There’ve been, on average, something like 2.5 fewer “bad” calls per game this season, thanks to the computers. As we gear up for the first postseason in which the system will thwart game-altering controversies, it’s a good time to take stock of the system as we’ve gotten to know it. Has it done what we expected it would do, for the people for whom we expected it to do it? Who’s gotten better (or worse) at using it as the season has gone along? And, crucially, because we live in an age defined by this question: How do we fix this new thing we just created?

I’ve had endless fun breaking down interesting case studies of ABS affecting individual players’ performance and value this year, from Cal Raleigh and Victor Caratini behind the plate to Junior Caminero and Alec Burleson at bat. Today, though, let’s stick to a few of the broader, systemic questions that come with the implementation of a new way of calling balls and strikes (or at least verifying them). To that end, I’ll start by answering the biggest question we should all have tracked, throughout the year: How much less common are bad calls, now?

Here, I’ve taken the percentage of all pitches not swung at that were classified by Statcast as being in the Shadow regions near the edges of the zone that were called wrong, in each direction. Another way to run this would have been by taking all pitches called a strike when technically outside the zone or a ball when technically within it, but in my opinion, anything outside the boundary that zone establishes is the responsibility of the umpires to clean up. We want to eliminate egregious misses, and the system should help with that, but that’s not the main job of the system; its main job is to assist umpires in what has become an insupportably difficult gig.