Time to Change it Up: An Examination of Changeup Usage and Effectiveness from 2015-2018

Research Team: Max Brill, Cam Cain, Theo Mackie

A version of this story is also available at MSABR.com.

This year in the Michigan Sabermetrics Club (M-SABR), we’ve made an effort to put a little more focus on the “R” in our name. This post is the culmination of a few weeks of research that a couple of my friends and I have been doing and it is very much a change of pace from what normally is published on this blog. If you’re not a math or stats person, this might be a tougher read than usual. I still tried to inject my usual dose of sarcasm and wittiness but if you decide that you didn’t enough after reading this article, stay tuned for next week.


A 2010 study conducted by Dave Allen over at The Baseball Analysts suggested that all else equal, pitchers should try to avoid throwing their changeups to same-handed hitters and should be more willing to throw their changeups to opposite-handed hitters. That is to say that right-handed pitchers should be throwing more changeups to left-handed hitters and left-handed pitchers to right-handed hitters. Here’s his reasoning:

If a pitcher can release his fastball and slider with roughly the same initial trajectory and locate his fastball around the middle of the zone the difference in movement will put his slider down and away to a same-handed batter and down-and-in to an opposite handed batter. If he does the same with his changeup the pitch will end up down and away to the opposite-handed batter and down and in to the same-handed batter. All else being equal a down-and-away pitch is much better than a down-and-in pitch.

– Dave Allen, “Musing on Pitch Type Platoon Splits” (2010)

2010 was nearly a decade ago and baseball statistics, technology, and data manipulation have come quite far since then, so we figured we would investigate Allen’s claim that a changeup thrown to an opposite-handed hitter should be more effective than one thrown to a same-handed hitter.

It has been well-documented that pitchers, generally speaking, have the platoon advantage when facing a hitter of the same handedness. The prevailing wisdom in baseball—based on Allen’s article—is that changeups should be thrown predominantly to opposite-handed hitters, not same-handed hitters. We actually found the reverse to be true: typical platoon splits tend to hold true, even when a pitcher throws a changeup. Here are our results.


For reference in the upcoming section, xBA is expected batting average, wOBA is weighted on-base average—it’s similar to slugging percentage except it uses different linear weights for different events and incorporates more offensive outcomes into the final number—and xwOBA is expected weighted on-base average.

The investigation into changeup platoon splits began by setting parameters for the baselines that pitchers would need to qualify for the study. Only pitchers had thrown a minimum of 50 changeups to right-handed hitters and left-handed hitters (At least 100 changeups total and a minimum of 50 to batters of each handedness) since 2015 were included in the data sets. Once this constraint was decided, four corresponding sets of data were downloaded from Baseball Savant using the criteria detailed above: RHP vs RHH, RHP vs LHH, LHP vs RHH, and LHP vs LHH.

We then calculated a handful of stats for each subset of data. A screenshot of our results is below. We calculated the median and the “mean” of each statistic, but labeling the value “means” was a little bit of a misnomer.

In the dataset, there was not enough data for us to compute wOBA, xwOBA, and xBA by hand, so we had to devise a way to come up with the means ourselves. Our first thought was to use the number of plate appearances each pitcher had and divide those by the total number of plate appearances within the data subset to get the percentage of the data set that each pitcher made up. Unfortunately, the data set didn’t include plate appearances, so we figured that using at-bats would be the next-best option.

After we calculated what percentage of at-bats each pitcher was responsible for in the datasest, we multiplied that number for each individual pitcher by that pitcher’s xBA, wOBA, xwOBA, etc. and then divided by the total number of at-bats to get a weighted average for each statistic. That is what we called the mean. I supposed “weighted average” could have been a more descriptive term, but “mean” gets the point across.

A quick example of our “mean” calculation in action: Adam Conley (a lefty) had 37 of the ABs against lefties in the selected group of 60 left-handed pitchers. We divided his 37 AB by the 2145 total AB that LHP had against lefty batters to get 1.7% (0.017): Conley’s contribution to the total weight of any statistic of which we calculated the mean. In other words, Conley’s totals in wOBA, xwOBA, etc. comprised 1.7% of the total wOBA, xwOBA, etc. that left-handed pitchers posted against left-handed hitters.

We ended up with 252 pitchers in the dataset for righties and 60 pitchers in the dataset for lefties.


Screen Shot 2018-11-09 at 1.29.02 AMScreen Shot 2018-11-09 at 1.29.38 AM

As you can see above, right-handed pitchers clearly fare better when throwing their changeup against right-handed hitters, which contradicts conventional wisdom about when changeups should be used. The same can be said for left-handed pitchers.

From 2015-18, right-handed pitchers were clearly more effective in nearly every way when throwing changeups to same-handed hitters as opposed to opposite-handed hitters. Over our four-year timespan, righty pitchers allowed a .225 BA against to righties and a .233 to lefties when throwing their changeups. The same goes for lefty changeups, though the gap was even more pronounced—right-handed hitters hit 52 points better against changeups from lefties over the past four years than left-handed hitters.

The data follows the same trend—that batters compiled a worse BA, xBA, wOBA, and xwOBA when facing same-handed pitchers—in all instances except for wOBA against right-handed pitchers. I’m not going to go through each category and each handedness because the data is all above, but you get the idea. What I will mention is the discrepancy in wOBA against right-handed pitchers.

As you can see from the table, right-handed hitters had a higher wOBA against right-handed pitchers during our timeframe than left-handed pitchers did (.300 to .291). The best explanation for this is luck and the fact that xwOBA, which takes launch angle and exit velocity into account, is in line with what is expected based on the rest of the data (same-handed hitters are at a disadvantage when facing changeups), points to batter wOBA versus right-handers simply being an outlier.

It is evident from the left-handed pitcher wOBA data that left-handed hitters performed worse against changeups from left-handed pitchers than changeups from right-handed pitchers (.270 to .300, a significant difference). That’s exactly what we expected. Even with the right-handed hitter wOBA discrepancy, the right-handed hitter xwOBA numbers fit our observation that right-handed hitters fared worse against right-handed changeups than left-handed changeups.

Speaking of xwOBA, let’s talk about exit velocity. Whereas the other metrics discussed so far have generally contradicted conventional wisdom about when it is appropriate to use changeups, the exit velocity numbers make no such distinction. Why am I including those numbers if they do not bolster my case? There are two main reasons:

  1. Transparency.
  2. Though the exit velocities alone do not tell us much, the exit velocity data coupled with the launch angle data tells us that pitchers were able to more frequently induce ground balls from same-handed batters than opposite-handed batters.

When it comes to launch angle and exit velocity, the second point is of the utmost importance within the context of this analysis.

Per Fangraphs, ground balls have the lowest wOBA of any of the three batted ball types—ground ball, line drive, and fly ball—so the more a pitcher can induce ground balls, the better off they will generally be. This is not groundbreaking stuff. What is somewhat groundbreaking is the fact that hitters have a worse launch angle when facing changeups from same-handed hitters.

On average, right-handed hitters facing changeups from right-handed pitchers had a launch angle 0.6 degrees lower than left-handed hitters facing the same group (8.30 launch angle for right-handed batters versus right-handed pitchers compared to 8.90 for left-handed batters). And once again, among left-handed hitters, the gap was even more pronounced: left-handed hitters hit the ball nearly three degrees lower when facing changeups from righties as compared to their right-handed hitting counterparts against the same sample of pitchers over the past four seasons.

The fact that the exit velocities are relatively similar for all matchups is, in short, irrelevant—if the ball is being hit on the ground more often by batters facing changeups from same-handed pitchers, the launch angle, not the exit velocity, is going to be affecting the batter’s wOBA more. This also explains the discrepancy between the actual wOBA and expected wOBA for hitters facing changeups from right-handed pitchers; the launch angle and exit velocity data points to the fact that right-handed pitchers should be faring better when throwing their changeups to right-handed hitters than left-handed hitters. As mentioned before, this was not the case in actuality—the results on the field did not match the expected results. That is, of course, part of the human element of the game and why we play baseball instead of simulating it on a computer.

That brings us to the final segment of our data: whiff rate and take rate. Though these data points are not results from balls put in play, the fact that the balls were not put in play tells us a lot about the effectiveness of pitchers throwing changeups to same-handed hitters.

We can see from the data above that, for the most part, the data about swings and takes is still in line with what we’d expect—batters whiffed more against changeups when they faced same-handed pitchers than when they face opposite-handed pitchers. This effect was more pronounced for batters facing right-handed pitchers than left-handed pitchers. There isn’t any obvious explanation for this, but the fact that the trend—that changeups generate more whiffs when thrown by pitchers who share a batter’s handedness—still exists in both instances is more evidence toward the fact that perhaps the conventional wisdom about changeups is wrong.

Batters also take fewer pitches against changeups from same-handed pitchers. This is partially correlated with the fact that batters swing and miss more on changeups from same-handed pitchers and, once again, the effect is more pronounced against right-handed pitchers than left-handed ones. Once again, there is no easy explanation for the fact that this effect is more pronounced against right-handed pitchers, but it still exists nonetheless.

Of course, all of this data needs context. It’s clear from the above analysis that changeups have been more effective over the past four years against same-handed hitters than opposite-handed hitters, but how much more effective have they been than other pitches? For this, we used the same methodology as described above, except instead of just using changeups, we took all non-changeup pitches thrown in 2018 and calculated the same metrics. We were unable to use the data from 2015-2018 simply because our computers could not handle the sheer size of the data, but we decided that the 2018 sample was both large enough and representative enough of the data over the past four years. Here’s what we found:

Screen Shot 2018-11-09 at 2.25.09 AMScreen Shot 2018-11-09 at 2.25.23 AM

From this data we can see that the general belief surrounding platoon splits is true: right-handed pitchers do better against right-handed hitters and left-handed pitchers perform better against left-handed hitters. This trend has been well-documented for years, so there’s nothing new there. What we can see from the data, though, is that the changeup is an incredibly effective pitch. Over the past four seasons, wOBA against changeups ranged from .270 to .300 in our sample. wOBA against non-changeups in 2018 ranged from .299 to .337. And this is not simply a product of “better hitting” in 2018; the league-average wOBA in 2018 was just .315, the fourth-lowest mark since 2000.

That is not to say that pitchers should drastically increase their changeup usage. It does, though, back up the fact that changeups should not be thrown only to same-handed hitters—in fact, pitchers could probably stand to increase their changeup usage to same-handed hitters based on the data presented in this article. Take left-handed pitchers throwing changeups to left-handed pitchers for example; there were only 2145 ABs that ended in a left-handed pitcher throwing a changeup to a left-handed hitter over the past four seasons even though that was, by the numbers, when pitchers were most effective throwing their changeups. On the other hand, right-handed pitchers ended 27595 at-bats against left-handed batters with changeups, the highest number of at-bats out of any of the four righty-lefty matchup combinations, despite that being the worst situation in which pitchers could throw changeups.

Nearly all of the trends that we found in our data for changeups holds true in the non-changeup data, so it’s not exactly clear why the prevailing philosophy in baseball is that changeups should not be thrown to same-handed hitters due to the fact that they move into, and not away from, a hitter. And there is certainly a compelling case that Dave Allen’s idea that a down-and-away pitch is much better than a down-and-in pitch is not necessarily true when it comes to changeups.


No study is perfect, so for us to publish this piece without acknowledging the flaws would simply be foolish.

The most glaring issue is the trouble we had with finding the mean of the rate stats that we used in our results. I think that the weighted average does a good enough job of estimating what the true mean would have been for each of those stats. Additionally, the sample size is large enough that if we were able to calculate the mean for those rate stats, I do not think that they would be too far off from the numbers we came up with through our weighted averages. Still, though, it’s worth mentioning that we would have been able to more accurately calculate a true mean if we had access to plate appearances, at-bat outcomes (singles, doubles, etc. as opposed to just “hits), etc.

Another issue that we briefly touched on was the sheer size of the non-changeup data. If we had the capability to, we would calculate the statistics we used for changeups with our non-changeup data dating back to 2015, but none of our computers could handle the size of that file. As I mentioned above, we thought that the 2018 data was both representative enough of the past four years and a large enough sample size to not make a significant difference in the numbers we crunched, so we did not pay it too much mind. If we had the resources to do this again, though, we would perhaps go a different route with calculating our numbers for non-changeups.

There is also the fact that within the changeup data, there have only been 2145 at-bats that ended with a lefty-versus-lefty changeup since 2015. That’s just 0.3% of the total at-bats since 2015, a minuscule amount relative to the size of the other matchup data we used within our changeup data set. Unfortunately, we cannot go back in time and tell left-handed pitchers to throw more changeups against left-handed hitters but if we could do it we would—both because it would give us more data points to work with and also because it appears, at least from our data, that left-handed pitchers should be throwing a lot more changeups to left-handed hitters. They are effective.


Though I’d like to believe that the implications of this study are massive, I’m well aware that plenty of people within baseball already knew about this trend prior to the publication of this piece. Still, though, I think that any piece that contradicts conventional wisdom, be it in baseball or any other field, is worthy of some discussion.

It is pretty clear from the above data that the conventional wisdom regarding changeup—that they should only be thrown to opposite-handed hitters—is nonsense. Platoon splits and the fact that pitchers tend to have the advantage when they are throwing to same-handed hitters has been well documented, so to think that that trend would change based on the fact that the changeup moves in a different direction than a slider or curveball does not make a whole lot of sense. The two-seam fastball, though its movement is typically not as pronounced as the changeup, is regularly thrown to same-handed and opposite-handed hitters alike. It is certainly a possibility that changeups are, in fact, more effective than the two-seamer. That’s a study for another time, though.

This analysis of the changeup reminds me, in some ways, about relief pitcher usage in baseball over the past forty years. After saves were invented, managers decided to artificially restrict the innings that their best relief pitcher could throw based on whether or not it was a save situation. In the case of changeups, which have been around for much longer, managers and pitching coaches have decided that changeup usage should be artificially restricted to opposite-handed batters only. Why? Simply because of the way the pitch moves. Our hope is that this study is, at the very least, able to jumpstart a discussion about changeup usage in baseball. And when if we conduct this study again four years down the road, I hope to see a higher concentration of same-handed changeups thrown.

Special thanks to Kyle Kumbier, Ryan Pinheiro, and Duncan Wallis for helping to facilitate this project.

Like what you read? Click the menu button in the top right corner and enter your email in the “follow” box. You can also follow Max on Twitter by clicking here.

One thought on “Time to Change it Up: An Examination of Changeup Usage and Effectiveness from 2015-2018

  1. Wow , comprehensive and impressive undertaking. I think this analysis clearly allows a conversation, further data analysis and a computer than can handle the data in the future will hopefully lend credence and validity to this trend. S


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s