Best of BP: Luis Castillo’s Disappearing Changeup

Image credit: © Steven Bisig-USA TODAY Sports

The more that I write about it, the more I’ve come to find that when pitchers lose a pitch, it often happens over the course of multiple seasons. Over the past several years, Luis Castillo hasn’t changed all that much, if we’re talking overall results. But when you look at how he’s dominating hitters nowadays, he only vaguely resembles the pitcher he was before. He’s scaffolded his way to this point, so to speak. 

You may remember that, in late 2017, Castillo developed a sinker and learned how to deploy his fastballs to play off of one another, and then in 2018 he began throwing his vaunted changeup more. But it wasn’t until 2019 when he started elevating his four-seam fastball when we saw him take his changeup and turn it into one of the best pitches in MLB. As far as his changeup goes, that was its peak. Natural regression was probably going to take place regardless of what he did, but that’s precisely the point: Castillo changed quite a bit after that year. 

In searching, it seems that no one wrote about Castillo’s changeup in 2020. In hindsight, maybe we should have, but then that’s exactly the point I made above. It’s not like his changeup was gone, but it wasn’t missing bats the same, and it was starting to fly over the fence more. It wasn’t until several months into 2021 that Malachi Hayes and Justin Choi both wrote him up on the same exact day, in which both hypothesized that changeup command was the culprit. Days before, I’d suggested to Choi that Castillo’s struggles were release-based, and that release issues were the mediating factor affecting his command. I still think it’s true that his release traits were behind his changeup’s declining efficacy, but perhaps not in the way that I would have described back then.

Release graphics are a great way to see visually where a pitcher is releasing the ball in the space, but there are some confounding variables like changes in arm angle and pitch extension that affect release height. A different way to capture how Castillo is releasing the ball is looking at the spin properties of his changeup itself. To do this, we’ll look at the expected and actual movement direction of Castillo’s changeup, as well as its observed and spin-based active spin percentages:

It wasn’t until 2020 that Hawkeye data became available, for the first time allowing us to directly measure spin axis and spin rather than inferring them from the observed path of the pitch. So, for 2019, we have the expected movement direction of Castillo’s changeup, but not the actual movement direction. The same goes for active spin. We have observed active spin going back to 2017, but not the more standardly used spin-based active spin. That means that we’re leaning on his observed active spin rate which, admittedly, is sketchy. 

With these numbers, we can see a few trends. Perhaps most important is the leap in expected movement direction from 2019 to 2020, and then a smaller one after 2021. The shift is upwards, contradictory to his release point which has lowered, giving us a good example of release point obfuscating underlying changes. Castillo’s release continues to lower, which you’d expect to lead to more side-to-side movement, but his repertoire as a whole has become increasingly more vertically oriented over the years, which isn’t to say that he’s throwing over the top now, but most of his pitches have lost the sink and run that typified his pitches. The lone holdout was his changeup, until 2022. 

At its peak, in 2019, Castillo was one of nine right-handed starting pitchers who has thrown a changeup with expected movement directions of 2:38 or lower. His changeup’s expected movement direction was like that of Félix Hernández and Stephen Strasburg. As his expected movement direction has shifted upwards, Castillo has added more active spin to it which, intuitively, makes sense: much like a fastball, it’s probably easier to throw a changeup more efficiently at higher expected movement directions.

The issue is that, in doing so, his changeup has lost the traits that made it special. Before, he was able to kill lift with a combination of its side-spinning profile and a healthy amount of gyro spin. The increasing of Magnus movement probably isn’t a good thing for a changeup like this, especially since it’s shifting upwards relative to before, but if I had to, I might wager that the loss of gyro spin has changed the orientation of the seams to no longer be optimal for seam-shifted wake to take place. As much as anything else, the increase in spin efficiency is the main reason that Castillo’s changeup has gotten worse, with the root cause being a change in finger pressure or the angle of his arm or wrist at release. But I’m happy to admit that, without the data, that’s conjectural at best. 

So, Castillo’s changeup. Regardless of what the culprit is, it’s gotten worse because of the way he’s releasing the ball. That’s informed not only the raw stuff of the pitch itself, but also his ability to locate it, and as a result, it’s not getting either in-zone whiffs nor chases as it once did. You can make the argument that he lost his changeup in any year following 2020, but it’s clear that that the last time he had that changeup was in 2019. Since then, it’s accrued the seventh-highest run value of any starting pitcher who’s thrown 1,000 or more changeups. It’s just not an effective pitch anymore.

Luckily, Castillo’s fastball and slider are now potent pitches for the exact reasons that his changeup is no longer. Castillo’s changeup now enters the zone as flat as it ever has, which isn’t necessarily a great trait, but it is for a four-seam fastball and gyro (or bullet) slider, which have both also flattened by vertical approach angle (VAA).

Consider the boring action of this four-seam fastball to Bo Bichette:

And then how his bullet slider plays off of his fastball perfectly:

Bichette has to look for 97, riding and running, but his slider comes in 14 mph slower than his fastball. In terms of movement, this slider does as close to nothing as possible, due to its gyro-heavy nature  Castillo gets exactly one inch of sweep, and less than an inch of induced vertical break. What makes it special is that it looks exactly like his fastball up through the hitter’s swing decision point, until it’s not, and it also enters the zone with a lot of synthetic sweep due to his low release, in spite of its lack of raw movement. 

Luis Castillo’s transformation can’t be understated. Years ago, he could essentially throw his changeup anywhere and all but ensure a whiff or weakly hit ball. Now, we can say largely the same about Castillo and his fastball. Given the news coming out of spring training, it sounds like Castillo is going to throw more sliders, which is good, because for every slider thrown, it means a changeup not thrown. That speaks to the demise of his changeup, but just as much to the emergence of his four-seam fastball and slider.