The Worst Pitches in Baseball This Year

Last Friday, Davy Andrews put Aaron Judge into context. That’s fun for the usual reasons – “Wow, look how dominant Aaron Judge is!” never gets old. But his final conclusion – Judge hits like most batters do when opponents hang a bad slider middle-middle – got me thinking. We have a pitch-level model that estimates the worst pitches. Can we use it to get an idea of what it looks like to throw a pitch so bad it turns your opponents into Judge-esque offensive producers?

There’s an easy way to sort this out. When you look at a pitcher’s player page here at FanGraphs, you can see how we model each pitch. There are a ton of scores, but I’m going to be focusing on PitchingBot today, for reasons I’ll explain shortly. The player pages break each pitch down by command and stuff. In our internal database, it’s even better: You can look at any individual pitch and get a grade on it. I set out to find the worst pitches on the year to see whether they made hitters look invincible.

Ironically, the worst pitch thrown this year made the batter look extremely vincible. Here it is in all its glory:

Yikes! PitchingBot assigns grades for each component, and it’s not hard to understand why this one got a poor command grade. If you only threw pitches like that, you’d never record an out. Poor James McCann — trust me, the ball hit the bat before he ducked out of the way — but that’s the only outcome that isn’t a disaster. Otherwise, it’s a ball or a hit batter.

All of the worst pitches this year are some variation of that, which tracks. Throw a cement mixer slider down the pipe, and there are still plenty of ways it can work out. The batter could take it, or foul it off. They could smash it directly to a fielder. Miss the strike zone by a matter of feet, and none of that is possible.

To get at the actual worst pitches in baseball this year, we’ll have to eliminate those obvious balls. New rules: The pitch has to be around the strike zone. The very worst pitches are all wild misses. I left those out and queried the model again. Here’s the new worst pitch in baseball this year:

Oh, right. We didn’t exclude position players pitching. Andrew Knizner’s pitch there graded out terribly on our stuff models, and rightfully so – 53 mph and right down the middle. But that’s not a real pitch, and I’m not even sure that the model knows what to do with an input that’s so unlike actual competitive offerings. Let’s add a new filter and only look at pitches thrown 70 mph or harder. Here’s the new worst pitch in baseball this year:

Wait, what? That pitch looked pretty good! What was Jesús Sánchez going to do with it? The reason it grades out so poorly is almost all down to location. Very few batters swing at pitches that are both high and off the plate inside. That one notwithstanding, it generally doesn’t work out well for the pitcher.

One thing I learned as I combed through these pitches is that when you drill down to a single pitch, command makes pitches much worse than stuff. It makes sense when you stop and think about it, but it still sounds wrong. Okay then, let’s recalibrate. We’re looking for pitches in the strike zone now.

Here’s the new set of criteria: We want pitches in the zone, thrown 70 mph or harder. I want to find meatball-y meatballs, pitches that are bad because they don’t break enough and are thrown too slowly. With that criteria sorted, we can finally get to some actually bad pitches. Here’s the new worst pitch of the year:

Yeah, that one fits. When you think of a duel between Justin Verlander and Nolan Arenado, you probably picture a ton of high fastballs getting fouled off or cut under. But sometimes Verlander just throws a clunker. That pitch had about half the horizontal break of a normal Verlander slider, with the movement instead translating to backspin. It fell five inches less than his average slider this year. In other words, it was a hanger. It had a terrible grade for both stuff and location. Sure, Arenado missed it, but you can look at that pitch and understand what pitching models are doing.

Here’s the second-worst pitch of the year, and wow, this one is a howler:

The model didn’t actually hate Caleb Thielbar’s location; in 0-0 counts, throwing the ball right down the middle isn’t a disaster. On the other hand, it hated the pitch itself. I see why. If you’re going to throw a slow curveball, you need more downward movement than that. By starting it so high, the pitch kind of lollipopped out of Thielbar’s hand, which made it fall less and with a less negative approach angle.

Here’s a weird fact: The nine worst pitches by this criteria all produced good results for the pitcher. There were four foul balls and five outs. But don’t worry, there are some pretty bad pitches that resulted in thunderously positive results for hitters:

This is a great example of location and stuff combining to make a pitch bad. That pitch wasn’t completely awful; it had roughly average movement relative to Nick Avila’s seasonal average. But in a 1-2 count, you can’t lob a slow pitch over the middle and expect the hitter to take. A bad time for a bad location on an average-ish pitch can still work out to “very bad” overall.

Let’s get one last example of that combination failure. Here’s Nick Anderson missing his target in an advantageous count:

There’s a lot to dislike here. The pitch was supposed to be off the plate away, but Anderson left it in a very hittable location instead. That was also an awful fastball. It fell five inches more than Anderson’s normal heater, right into Jordan Westburg’s bat path. It ended up as straight as can be – it had less tail than his regular offering and was three ticks slower. That’s just a tremendously bad fastball, and in a bad location given the count. Maybe you can get away with missing one early in the count when a hitter might be looking elsewhere. With two strikes, Westburg was always going to swing at it, and this is not the kind of pitch you want a batter swinging at.

In the end, that’s capturing a lot of what Davy was looking at when he compared hanging sweepers to Judge’s overall offensive production. The worst pitches in baseball this year, as measured by our stuff model, are mostly balls. But if you strip those out, and remove position players pitching while you’re at it, you’re left with combination failures, pitches that hung up in dangerous locations.

What’s the point of all of this? To some extent, I’m just enjoying asking the model open-ended questions like “Which pitch do you like least?” and seeing what it says. I’m also curious, though. What does a bad pitch look like in the eyes of the computer? Luckily, a lot of it is fairly intuitive. Hanging breaking balls, slow fastballs that don’t have enough ride, and wildly missed locations. A pitching coach could have told you that years ago. Now, we just have the data to say the same. And of course, Aaron Judge would wallop these pitches, too, just so we’re clear.