Last Year’s Unluckiest Changeup

In baseball, luck is a tricky concept. In some cases, it’s used to describe an event that’s within the normal distribution of outcomes but far from the mean. In other cases, what we call luck might actually be the first signs of an outlying skill for which we simply lack a sufficiently large sample to identify.

We’ve developed a new understanding on one kind of luck in recent years — namely, the sort that occurs with a batted ball. With Statcast data, we can look at the shape and size of a ball in play and try to decide what the batter “deserved” from that sort of ball in play. Then we compare it to actual outcomes. The difference between the observed and expected outcome is luck.

What if you want to look at a luck on a specific pitch type, though? How would you do it? You could look at the results on the pitch and basically use the Statcast-type process from the other side of the ball. What sorts of balls in play did that pitch produce, and what sort of results should those balls in play have produced? The problem with that approach is that you’re slicing a pitcher’s repertoire into small samples when you start talking about balls in play off a specific pitch. Even David Price, for example — who led the majors in innings last year — allowed fewer than 300 balls in play on his most frequently thrown pitch, the fastball. Secondary pitches are, almost by definition, thrown much less often. Variance isn’t the exception in such cases, but the rule.

So I propose a different method for evaluating luck on specific pitches. For changeups, we know that drop, fade, and velocity differential are good for the pitch. And we can judge actual outcomes using whiffs and grounders — the former allows us to consider all balls in play at least. If we compare how good the changeup should have been based on movement to how good it was, based on results, we can get a sense of the luckiest and unluckiest changeups in baseball. It’s a similar process, but it avoids some of the pitfalls. Let’s try it out.

First, the fun leaderboard that shows us the best changeups last year by movement and velocity. The weakness in the analysis here is that all three aspects of the changeup are weighted equally, while some aspects of movement may be more important than others. For now, using equal weights of the three variables identifies these changeups as the best in the game.

Last Year’s Best Changeups by Movement & Velo
Pitcher swSTR% GB% Diff velo ABS Diff X Diff Z MoveVeloZ
Scott Kazmir 17.0% 28.8% 14.8 3.7 7.6 4.0
Brad Boxberger 12.3% 33.3% 12.8 4.9 8.1 3.9
Tony Sipp 27.9% 26.9% 13.3 3.1 8.3 3.4
Jharel Cotton 17.7% 35.0% 15.4 2.4 7.1 3.4
Fernando Rodney 21.4% 21.6% 12.4 5.3 6.6 3.2
Blaine Hardy 22.2% 24.0% 9.5 6.8 6.6 2.6
Michael Fulmer 19.0% 47.1% 8.9 4.8 8.3 2.3
Daniel Mengden 6.6% 33.3% 10.9 4.5 6.7 2.2
Brad Brach 20.5% 38.3% 8.3 4.4 8.2 1.7
Christopher Devenski 21.3% 32.7% 11.4 1.6 7.7 1.5
Drew Storen 15.0% 31.4% 7.7 3.7 8.8 1.4
Carlos Rodon 12.4% 31.3% 9.9 3.6 6.9 1.4
Julio Urias 19.0% 27.0% 12.6 5.2 2.8 1.3
Tommy Kahnle 13.9% 35.7% 7.8 6.8 5.2 1.2
Tanner Roark 14.9% 40.8% 8.6 6.6 4.7 1.1
A.J. Ramos 23.5% 40.6% 6.7 5.6 7.2 1.1
Fernando Abad 14.3% 29.4% 18.0 2.7 0.0 1.1
Mychal Givens 17.9% 50.0% 8.8 2.6 8.3 1.1
Roberto Osuna 13.3% 27.3% 12.7 4.4 2.8 1.0
Adam Warren 14.5% 44.4% 8.3 5.6 5.6 1.0
Movement & velo defined off of four-seamer if the pitcher had one; Average difference between change and fastball, for right-handers: 2.5x, 4.8z, 8.7mph

Generally, these changeups produced excellent results, as well. Noting that 14% represents a general “average” changeup whiff rate, you see 17 of the top 20 here are average or better. Six of these changeups are elite by whiffs. Four or five manage the feat of having good whiffs with a good grounder rate.

It’s heartening to see some less established changeups on this list along side ones we’ve known and loved. Scott Kazmir, Fernando Rodney, A.J. Ramos, and Tony Sipp all have changeups that have driven their careers so far. But Michael Fulmer, Jharel Cotton, Christopher Devenski, Carlos Rodon, and Julio Urias are younger pitchers with more question marks. But. If you’ve watched them pitch, you know: they have very good changeups.

Let’s now bring results into the equation. By doubling whiff rate z-scores and adding in z-scores for ground-ball rates, I created an easy metric for pitch quality that does decently when tested by ERA. I simply subtracted the results z-scores from the movement z-scores to find the pitches that did well by movement but not by results.

Last Year’s Unluckiest Changeups by Movement & Velo
Pitcher swSTr GB% Diff velo ABS Diff X Diff Z Results Z MoveVeloZ Diff from Results
Daniel Mengden 6.6% 33.3% 10.9 4.5 6.7 -3.2 2.9 6.1
Brad Boxberger 12.3% 33.3% 12.8 4.9 8.1 -1.1 4.7 5.8
Scott Kazmir 17.0% 28.8% 14.8 3.7 7.6 0.1 4.7 4.7
Mike Wright 8.2% 30.8% 13.2 2.8 3.7 -2.9 1.6 4.5
Tyler Wilson 5.6% 42.1% 8.2 5.4 5.5 -2.7 1.6 4.2
Bartolo Colon 12.1% 7.4% 8.8 3.6 3.9 -4.0 0.2 4.2
Tyler Duffey 8.4% 16.1% 8.1 3.8 3.6 -4.4 -0.2 4.2
Yovani Gallardo 3.3% 29.3% 6.0 5.1 2.6 -4.9 -0.9 3.9
Zach Eflin 1.3% 30.0% 6.5 2.5 3.2 -5.5 -1.6 3.9
Homer Bailey 6.5% 12.5% 6.4 1.4 4.1 -5.5 -1.8 3.7
Carlos Rodon 12.4% 31.3% 9.9 3.6 6.9 -1.3 2.2 3.5
Tyler Clippard 11.5% 25.7% 9.4 -2.2 10.9 -2.3 1.1 3.4
Luis Severino 8.9% 26.1% 7.3 5.0 3.8 -3.1 0.2 3.3
Roberto Osuna 13.3% 27.3% 12.7 4.4 2.8 -1.4 1.8 3.3
Joe Ross 10.1% 19.5% 6.2 5.8 3.3 -3.4 -0.2 3.3
Results Z = two times whiff rate z score plus GB z score
Movement & velo defined off of four-seamer if the pitcher had one

Kazmir was the only pitcher here to produce above-average results with his changeup, but his league-leading movement and velocity difference suggest he should have had better results. The rest of these guys all had poor results on their pitches and may have deserved better.

Perhaps we shouldn’t look too closely at Yovani Gallardo, Zach Eflin, and Homer Bailey. Their changeups looked bad by movement and velocity, so we’re just saying their results should have been less bad. I wouldn’t run to Vegas with those names on your betting slips.

The class that sits in between, the guys with average-looking changeups that did poorly on results, those are harder to figure. Maybe Luis Severino and Joe Ross and Tyler Duffey get incompletes on their report cards. They could go either way.

But there are a few pitchers here that have outstanding pitches by movement and velocity who had horrid results. Brad Boxberger can be better, Carlos Rodon seems to be headed for a breakout, and maybe Roberto Osuna will trust his changeup a little more if he gets good results early on next year.

The unluckiest changeup, though? Looks like it belongs to a decently unlucky guy in the green and gold. It looks like this:

Daniel Mengden fell to the fourth round after his velocity fell due to a stress fracture in his last year in college — from putting on too much weight in the weight room, he told me — and when he finally got to go back to his old funky delivery and put in a full year in Oakland, he ran out of gas late last year. Then he hurt his foot throwing a bullpen at home this offseason.

Surgery means we won’t see Mengden to start the season, but given how teams use starters these days, we will probably see him, and his changeup, sometime this year. Maybe then he’ll get results that link up better with the quality of his change-pieces.





With a phone full of pictures of pitchers' fingers, strange beers, and his two toddler sons, Eno Sarris can be found at the ballpark or a brewery most days. Read him here, writing about the A's or Giants at The Athletic, or about beer at October. Follow him on Twitter @enosarris if you can handle the sandwiches and inanity.

11 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Jimbomember
7 years ago

I understand you’re trying to simplify, but doesn’t this approach ignore mechanics? As in, if a changeup has good movement, but poor results maybe it’s because the pitcher is slowing his arm down and not “selling” the pitch or something similar.

Zane Jakobs
7 years ago
Reply to  Eno Sarris

I think we could get decent results by looking at release point difference between the fastball and changeup. If the pitcher’s arm speed is changing, chances are his release point will too.

Cheeknbut
7 years ago
Reply to  Eno Sarris

The people need to know, what was the luckiest changeup?