Filed under:

# Checking in on Rays changeups

What’s the state of the pitch?

Two summers ago, I explored how the shape of the changeups that the Rays threw influenced the groundball rate they saw on these pitches. Using a simple model that regressed vertical movement, horizontal movement, velocity and the velocity difference between each pitcher’s changeup and fastball against their groundball rate, I was able to identify which components of changeups were important in generating groundballs, and to construct an “expected groundball rate” for each pitcher based on the shape of their changeup.

Since the Rays leaderboard in that study is populated with names like Ronald Belisario, Kevin Jepsen and Nathan Karns, a refresher seems like a good idea. The model itself could also use some updating, but I’m confident in it to give us a general idea of how each pitcher’s changeup is likely to perform. If you’d like to read more about the motivations behind this and the details of the model, those can be found here. If not, the math discussion is over so read on.

This season, the Rays have a relatively low team groundball rate, at 42%, despite throwing the eighth most changeups in the league. But, this doesn’t necessarily mean that the Rays are throwing a ton of ineffective changeups—it instead highlights an overall theme of pitchers generating a lot of fly balls, a product of rising fastballs and pitching high in the zone. While the Rays certainly are not relying on keeping the ball on the ground to the same extent as the San Diego Padres or Colorado Rockies, groundballs can still be helpful in inducing weak contact and generating double plays. Below is a table with each Rays’ pitcher’s changeup measurements and their expected groundball rate on the pitch.

### Rays 2017 Changeup Expected Groundball Rate

Name Velo Velodiff H mov V Mov xGB%
Name Velo Velodiff H mov V Mov xGB%
Matt Andriese 86.36 6.09 -2.32 -0.13 62.32%
Chih-Wei Hu 88.49 5.10 -10.25 2.34 59.70%
Ryne Stanek 89.58 9.12 -5.70 2.84 55.27%
Adam Kolarek 75.97 12.79 9.41 -2.69 55.22%
Jose De Leon 83.94 7.61 -11.25 2.39 54.79%
Sergio Romo 81.58 5.69 -10.38 3.26 53.88%
Chase Whitley 83.76 7.32 -7.65 3.35 53.17%
Jose Alvarado 98.59 9.77 2.17 6.37 52.46%
Jake Odorizzi 84.09 7.63 -6.49 4.42 50.99%
Austin Pruitt 85.54 6.74 -4.75 5.29 50.97%
Chris Archer 86.21 9.52 -7.22 4.45 50.11%
Erasmo Ramirez 84.03 8.69 -8.26 4.56 49.63%
Alex Cobb 85.71 5.97 -7.57 6.47 49.58%
Jacob Faria 81.25 10.99 -5.60 3.51 47.92%
Brad Boxberger 81.17 12.48 -8.89 2.89 47.56%
Ryan Garton 85.54 7.76 -5.28 9.31 42.31%
Blake Snell 85.71 8.58 7.22 9.35 41.50%

While many names in this table have groundball rate above the league average on changeups of 49.6%, the pitchers who have logged the bulk of the innings this season are at or below league average, dragging the overall average down. Focusing on three pitchers in particular, xGB% gives us a jumping off point for more in-depth analysis.

## Blake Snell

During Blake Snell’s ascent to the majors, his changeup consistently received high praise from reputable sources, like FanGraphs and Baseball America. In his first big league appearance in 2015, it looked like it was going to live up to expectations. It had drop and above average run, albeit in a sample of three pitches. Since then, the changeup has been trending in the wrong direction. While Snell has added velocity to it every year, he has increased vertical movement (and decreased drop) substantially, which is negatively correlated with groundball rate.

According to this model, Snell’s changeup should have the lowest xGB% on the team, and excluding Ryan Garton, it isn’t particularly close. Not only is his xGB% well below league average, he has posted a 37.5% groundball rate and a 23.3% whiff rate on the pitch, which are both well below league average in their respective categories.

But, the pitch has being serviceable in the past, and maybe that minors labeling wasn’t as off-base as this year’s performance makes it seem. Plugging in the measurements from Snell’s 2015 appearance, we would expect him to generate groundballs on 45.68% of his changeups put in play—still below league average, but not nearly as low as this year has been.

It also looks like some of the struggles have come from his pitch locations as well. From the below heat map of his changeup, we see that he has thrown a significant number of them in the strike zone. Odds are this is a product of poor command, but the issue is likely multifaceted.

Had Snell thrown enough innings to qualify, he would have the lowest first strike rate in the league this season. As Brian Anderson pointed out on Saturday’s broadcast, Snell falling behind in the count so often makes it harder for him to go to his secondary pitches, and if he does, harder for him to be effective with them. Looking at Snell’s pitch breakdown in 0-0 and 1-0 counts, we can get a picture of how he approaches hitters when falling behind. On the right is a breakdown of Snell’s pitches on 0-0 and on the left is a breakdown on 1-0, against right handed hitters.

After falling behind 1-0, Snell either goes back to his fastball or throws his changeup, and in order to avoid moving to 2-0, Snell has to force these pitches over the plate. Instead of nibbling at the edges and staying out of dangerous areas, Snell is leaving his changeup vulnerable for opposing hitters to take advantage of. But, in the at bats in which Snell gets ahead 0-1, he has much more flexibility in what he can throw, as he can bury a slider in addition to throwing his changeup and fastball, as demonstrated below.

While I believe in general that the idea that the first pitch is most indicative of an at bat is often overstated, in this case, it gives some explanation as to why Snell’s changeup has struggled to get results. Hopefully, the issue resolves itself as Snell develops better command at the major league level.

With a slider and a curveball to complement an electric fastball, it’d be easy to suggest that Snell simply scrap his changeup and continue to develop his other off-speed pitches. However, Snell needs his changeup. He leans on it heavily against right-handed hitters, against whom his slider and curveball are less effective. Snell clearly had this pitch working at some point in his career, but now it is a matter of either finding that again or ironing out the changeup he is throwing now.

## Chris Archer

Toward the bottom of the list, but still above league average, is Chris Archer. Archer’s changeup has clearly been his third pitch throughout his career, with his fastball and devastating slider taking the main stage. But, it has now established itself as a pitch that can generate groundballs at a league average rate, and has the upside to turn into much more.

Over the last few years, Archer has been adding drop, or subtracting vertical movement, from his changeup. Vertical movement is strongly negatively correlated with groundball rate for changeups, so it makes sense that Archer’s changeup’s groundball rate has been rising since his entrance into the league in 2013.

But, Archer has been using the pitch differently this year. In the past, Archer has thrown his changeup primarily early in the count against left-handed hitters, whom his slider is less effective against, but also mixed it in against right-handed hitters on occasion. This year, however, he has yet to throw a changeup to a right handed hitter. While the sample of changeups put in play against RHH and LHH is incredibly small, two of the seven changeups put in play last year against RHH went for home runs. By cutting out these pitches, Archer has improved his overall performance on the pitch.

#### Two Changeups?

Last February, I proposed that the one reason that Archer was able to be so successful with a small pitch arsenal was him throwing multiple “versions” of his slider, effectively expanding his arsenal. Ian Malinowski flushed this idea out at the start of the season, where differences in velocity on his slider provided stronger evidence that Archer was throwing more than one type.

This season, it looks like Chris Archer is not only throwing two versions of his slider, there seems to be some evidence that he is throwing two versions of his changeup as well. I don’t think I’m ready to conclude that there are definitely two pitches, but it seems like there’s something going on here.

Looking at Kernel Density Estimations of Archer’s changeup’s movement this season, we see two peaks in the distribution. I prefer using KDE to look at the distribution of the shapes because it eliminates the ambiguity of the bin sizes in a histogram, but they essentially show the same thing. Below are KDEs for Archer’s 2016 and 2017 changeup movement.

While these graphs may look convincing, there’s a lot more to the puzzle that likely warrants its own piece. First, two peaks may stem from differences in recording measurements. Perhaps the Statcast cameras at the Trop are calibrated slightly different, so Archer gets an extra inch or two of movement during his home starts compared to his away starts. Or, Archer might have made an adjustment somewhere in the middle of the season in how he throws the pitch. Either way, it is something to watch as the season progresses and we gather more data from different parks.

At worst, it’s a league average pitch that complements two excellent ones, and Archer’s limited usage of it shields it from being exposed as anything more than that. At best, Archer is throwing two versions of it and keeping left handed hitters off balance, further diversifying his arsenal.

## Matt Andriese

At the top of the leaderboard is Matt Andriese, whose 62.32% xGB% leads the Rays by a healthy margin. This shouldn’t come as a surprise—Andriese’s changeup is sixth in the league in groundball rate this season, sandwiched between that of Corey Kluber and Max Scherzer in the category. While often classified as a changeup, Jason Collette has pointed out that this pitch is actually another instance of a Rays’ pitcher throwing a pitch that is a “hybrid” that doesn’t fit cleanly into a specific category. Andriese’s grip on the pitch is somewhere in between a changeup and a cutter, which gives it slightly less horizontal movement but devastating vertical movement, and led Collette to label it a “cut-up”.

Furthermore, the movement on this pitch has been astonishingly consistent this season. A while ago, I wrote about how looking at histograms of movement shows the distributions behind the averages, and can expose a seemingly average pitch that is actually devastating some of the time and ineffective at other times.

Looking at vertical movement, he has a tight distribution on the pitch. Of the 292 changeups he has thrown this year, only four have had more vertical movement (less drop) than league average.

Looking at the histogram of his horizontal movement, we see that it is pretty tightly distributed with a slight tail to the left. The consistency of this shape is encouraging, as Andriese can be pretty sure how it is going to move, and in the case that it behaves differently, it moves more than he would expect – it’d be more worrisome if the pitch had a tendency to have no movement, or movement in the other direction if executed imperfectly.

While the execution of the pitch has been excellent, on the surface, the Statcast measurements on this pitch don’t look extraordinary. In 2017, hitters are putting Andriese’s changeup into play at a launch angle of 1.73 degrees, which is much lower than the league average of 8 degrees. But, the exit velocity on Andriese’s changeups is slightly higher than league average, at 85.5 mph compared to a league average of 84.8 mph.

While these stats are often used in tandem, their marginal impact on the result of the ball in play isn’t always the same. Meaning, a decrease in exit velocity from 95 to 90 mph isn’t going to decrease the wOBA on a ball in play, for example, the same as a decrease from 85 mph to 80 mph would. Furthermore, exit velocity is more important than launch angle on some balls in play, but launch angle is more important in others. This point can be illustrated by looking at the wOBA on balls in play for each launch angle and exit velocity.

Here, we see that wOBA is pretty consistent from 70-90 mph, but really starts to increase as the exit velocity passes 95 mph. For launch angle, it rises pretty steadily at -7 degrees, peaks somewhere around 20, and then drops quickly after that. In the case of Andriese, we would expect his changeup performance to be similar whether it had been put in play at 80 mph, 85 mph or 90 mph. But, Andriese’s ability to decrease the launch angle on the pitch greatly improves the results he is seeing on it. A decrease in launch angle from 8 degrees to 1 degree is somewhere around 200-250 points of wOBA, or roughly the difference between Mike Trout and Darwin Barney’s 2017 seasons.

With an effective, consistent shape that minimizes the launch angle on the pitch, Matt Andriese’s changeup had cemented itself as one of the best in the game. Andriese isn’t eligible to return from the disabled list until August, so all we can do is hope that he picks up right where he left off during the final stretch of the season and effectively be another trade deadline acquisition.

Data from Brooks Baseball, Fangraphs and Baseball Savant.

Editorial note: Last night Jesus Sucre pitched an inning. It’s hard to tell what’s a changeup and what’s a fastball when the average velocity is 81 mph, so we aren’t really going to try. But if you’re craving analysis of the pitch, here it is: Not Good. No ground balls produced so far, and unlikely to do so at a high rate going forward.