Yesterday, I put every fastball thrown in baseball last year into a giant spreadsheet to come up with expected pitch values. Well fine, it was a small snippet of code, not a giant spreadsheet. But the output came in a giant spreadsheet! In any case, the idea is pretty straightforward: look at a player’s pitches, substitute in xwOBA-based contact numbers instead of actual results, and call it a metric.
Today, I’m completing the set. Well, I’m kind of completing the set; I ignored knuckleballs because there aren’t enough of them, and secondary offerings are more complex. Due to differing classification systems, I scraped breaking balls (sliders, curveballs, knuckle curves, and even cutters) and offspeed pitches as a single pitch type. Otherwise, we might end up with something like Nick Anderson — classification systems can’t decide if he throws a curve or a slider.
One more thing: the system is heartless. No human could argue that this wasn’t the best curveball of the year:
Or if not that one, then this one, with bonus Eric Lauer bewilderment and Greinke sprinting:
Slow curves are undoubtedly the best curves, results be damned. But the soulless calculation…