Statistical Signature of a True Changeup Artist

Cole Hamels is one of baseball's best changeup artists. (via HyunJae Park)

Cole Hamels is one of baseball’s best changeup artists. (via HyunJae Park)

I’ve never quite understood the concept of a change-up “speeding up the bat” of the opposing hitter, since in my mind, a change-up would give the batter incentive to delay his swing a little, effectively slowing down the bat. Further, baseball mechanics are designed to be repeatable and are largely just muscle memory (much like a golf swing), so it is highly unlikely that the batter’s actual bat speed will be increased.

I bring this up since quite often we hear baseball people, usually of the “non-analytical” subset of baseball fans, say things that don’t always make sense on the surface, but actually have some underlying truth. Prime examples of this are the analytics view on catcher defense and non-FIP pitcher ability, where we began by saying they don’t exist and can now point to measurable effects to varying degrees (catcher framing, velocity and preventing homers, etc). Further, I am loathe to dismiss any observations made by someone who’s experienced the game for over a decade at the major league level.

Are we able to find any evidence to support the theory that change-ups can “speed up the bat” of the opposing hitter? If so, is there an inherent cost to throwing a change-up that is hidden in subsequent pitches? Do we need to adjust the pitch values (such as those published on FanGraphs) to include the diminished effect of future pitches? Alternatively, is there a specific effect that is dependent on what type of pitcher is throwing the change-up? Let’s take a deeper look and see what the data say.

Below is a graph that compares two specific pitch combos, FF-CH and CH-FF. FF-CH indicates a four-seam fastball followed by a change-up in the same at bat. The reverse is true for CH-FF, which would be a four-seam fastball thrown after the change-up. We’ll contrast that with just FFs and CHs and see if there is any difference in outcome, specifically because of the previous pitch.

Pitch Outcomes CH-FF to just CH

The green bars represents a change-up thrown, the blue bars a four-seam fastball. On the left, these are pitches that are preceded by a FF or a CH specifically, on the right the pitches are preceded by any pitch type. What you’ll notice is how the data are very consistent for many of the outcomes (swing percentage, foul percentage and groundball percentage) where throwing a change-up first doesn’t affect the following pitch much. They differ in two specific ways, which appear to be telling the same story:

  • Throwing a CH-FF as opposed to Anything-CH will get you two percentage points more of called balls, two points fewer called strikes and two points fewer swinging strikes. This is a significantly bigger impact than the FF-CH combination, which mostly impacts called balls. Given that swing percent is stable and that the average distance from the center of the strike zone does not change (i.e. the location of the pitch is equivalent) it would appear that umpires are slightly fooled when they see a different pitch type and are slightly more likely to call something a ball.On the flip side, only the CH-FF combo had a significant impact on the called and swinging strikes, lending credence to the notion that hitters have a “quicker” bat as evidenced by their reduced propensity to swing and miss and reduced propensity to take a called strike (quicker bats allow the hitter to wait longer to get a better gauge of whether a pitch will be a called strike).
  • Slugging percentage and home runs are boosted in both combinations, but significantly more in the CH-FF scenario. SLG increases by 15 points in the CH-FF and by 10 points in the FF-CH combo. HRs are significantly more likely post-change-up than post-four-seam. My working theory is that seeing one pitch type or the other allows the hitter to time the pitches better by giving him more information about which pitch type he is seeing. The other half of my theory is that it is easier to see that a pitch is moving faster than it is to see than it is to see that it is moving slower, thus augmenting this effect in the CH-FF combination.

What happens when we throw the same pitch back to back; i.e., is a CH-CH better than anything-CH?

Pitch Sequencing Graph 02

For this contrast, I’m going to point out just two things, the whiff rate for CH-CH and the swing percentage for CH-CH. Notice how the swing rate ticks up in the CH-CH combo, but not in the FF-FF combo; this is likely due to hitters (incorrectly) guessing that a pitcher isn’t going to throw back-to-back change-ups, thus leading to an increased propensity to swing as well as an increased propensity to swing and miss when that pitch is a change-up. Based on these data, it looks like throwing two change-ups in a row is actually a pretty good strategy.

What about a change-up artist like Cole Hamels.

Seq_03_hamels

Look at the HR/Swing for the CH-FF combo! The average Hamels four-seam fastball will leave the yard 1.6 percent of the time when swung at, but when it is thrown after a change-up, this jumps up all the way to 2.3 percent. You can see this represented in the SLGContact variable as well. However, note how Hamels also gets an extremely high swing rate on the CH-CH, despite worse balls-in-play performance (fewer home runs, lower slugging) on the CH-CH combo versus the average change-up. It appears that while Hamels’ fastball becomes significantly less effective after a change-up, it actually improves his change-up.

Let’s see of this trend is repeated with another change-up specialist. Here’s Marco Estrada:

Seq_04_estrada

This same effect is present after Estrada’s change-up, with CH-CH producing much better overall contact management, and CH-FF producing much much worse contact management and causing a three percent HR/Swing rate. So we’re beginning to see a pattern here, which mostly affects the quality of contact of the flour-seamer when it follows a change-up (at least for classic change-up guys). Let’s look at just slugging for a wider group of pitchers, specifically the top change-up throwers by quantity of change-ups thrown since 2008:

Seq_05

Quickly jumping out are three pitchers who suffer from the post-change-up effect, Bronson Arroyo, Edinson Volquez and Jeremy Guthrie, with fastballs that all become extremely hittable post-change-up, especially in comparison to a fastball thrown after a fastball. In stark contrast are Tim Lincecum, John Danks, Francisco Liriano and Jeremy Hellickson, who appear to have the reverse effect. James Shields and CC Sabathia are relatively neutral from this standpoint.

This brings us to another, possibly more intriguing, observation:

Hitters may be able to pick up when a pitch is the same, rather than when a pitch is different. Further, we can assess the quality of a change-up’s ability to be picked up by the hitter by the quality of contact on the CH-CH combo.

A Hardball Times Update
Goodbye for now.

This is founded on the theory that a Hamels or Shields change-up, which can completely fool the opponent into thinking it’s a fastball, is far less likely to be picked up by a hitter as a change-up, even when thrown back to back. This would be a signal of a pitch that can be thrown in any count. The key indicator here is where CH-CH produces a lower SLGContact rate than something-CH.

Contrast that to a Felix Hertnandez change-up or that of a lesser pitcher such as Hellickson, where hitters jump all over the CH-CH combo, clearly picking up something when they see it twice in a row. This would lead me to believe that Justin Verlander has quite a deceptive change-up, whereas Lincecum does not. This can then separate the men from the boys: A true change-up artist can throw a change-up even after he’s already thrown one; a lesser pitcher can throw it only after a fastball, or another pitch type. This does not appear to be a velocity-based requirement either, since Jered Weaver has also shown the ability to throw back-to-back change-ups.

Conclusion

Sequencing has a very clear and consistent, pitcher-specific effect on a pitcher’s ability to throw a subsequent pitch type. While certain pitchers may have change-ups that are great when they follow a fastball, the true change-up artist will be able to throw back-to-back change-ups without any adverse effect. This could help explain why pitchers who have great swinging strike rates on the change-up can often get hit really hard. Pitchers who can throw a good change-up, but are susceptible to hard contact on back to back change-ups will basically lose a pitch when they throw a change-up, making them much more hittable.

A pitcher who is a true change-up artist can throw the CH-CH combo, but beware that his change-ups will likely tee up his fastballs a little.


Eli Ben-Porat is a Senior Manager of Reporting & Analytics for Rogers Communications. The views and opinions expressed herein are his own. He builds data visualizations in Tableau, and builds baseball data in Rust. Follow him on Twitter @EliBenPorat, however you may be subjected to (polite) Canadian politics.
13 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Billy H
8 years ago

“Speeding up the bat” is not used literally by most baseball minds. Rather, it is referring to throwing a slower pitch when the hitter can’t catch up with your hard stuff. If you just threw a fastball by him, the classical thinking goes, you should stick with the fastball since he was too “slow”. Throwing him a changeup would make his late timing better, giving him a better chance to hit the ball.

Matt K
8 years ago
Reply to  Billy H

Exactly. Its not actually speeding up his bat, its slowing down to his bat speed with a softer pitch. This “effectively” speeds up the bat.

Also I would like to see the amount of CH-CH each pitcher threw. I would imagine that the totals very greatly depending on the effectiveness/confidence a pitcher has in that pitch. I would imagine the effects measured here and the total amount of times a pitcher doubled up on a CH would be highly correlated.

Peter Jensen
8 years ago

Wouldn’t your results on the second pitch change quite a bit depending on whether the first pitch was a strike or a ball and also on the counts in which they were thrown?

baseballfan123
8 years ago

“Throwing a CH-FF as opposed to Anything-CH will get you two percentage points more of called balls, two points fewer called strikes and two points fewer swinging strikes. ” This is a classic error of correlation does not imply causation. Just because the numbers tell you that it does not mean that throwing a FF or a breaking ball will “get you” 2 percentage points. Maybe the real reason is that pitchers that tend to follow that pattern have higher Ball rates, etc

Zovat Andros
8 years ago
Reply to  baseballfan123

Yeah, exactly…or maybe those combinations are thrown in slightly different counts and therefore more to the edges of the plate. Or maybe we’re comparing a whole bunch of power pitchers throwing breaking balls and fastballs in the “anything-FB” category to a smaller number of finesse guys in the CH-FF category, which would explain why the anything-FB is more over the plate and more swing-and miss. There’s a lot of this in this article — analyzing pitch sequencing is not so simple as lumping everything together and drawing conclusions as if all pitches are thrown in the same environment.

(I also think that’s a typo and he means “CH-FF as opposed to anything-FF” or else the numbers don’t match up to the statements and I’m not sure why those two are being compared anyway.)

Bad to the Zone
8 years ago

It is a shame this technology was not around when Gene Garber used his FASTBALL as a CHANGEUP!
And what was it I read recently about pitcher’s who throw a changeup having a higher than average FIP?

https://www.youtube.com/watch?v=ZgubG-MOPT4

channelclemente
8 years ago

Although it sounds like carping on trivia, if you’re going to consider FB’s in sequences with CH’s, I’d either focus on the 2 seam FB, or at least split it out for separate consideration. It is a more deceptive offering in such a sequence since it follows a similar trajectory to most CH’s thrown with the 2 seam grip.

Ben
8 years ago

I would think that a disproportionately large portion of the CH-FB scenarios would be a pitcher needing a strike after missing with a CH putting himself behind in the count. Those get-me-over FB’s have a much higher tendency to get crushed.

Jetsy Extrano
8 years ago
Reply to  Eli Ben-Porat

Or giving a clue that Hamels manages this style because he has better CH command, which also makes him a good pitcher?

steve breault
8 years ago

I’m curious to know if you pull all PitchFX and/or MLB data yourself or do you use a web service?

John
8 years ago

Hopefully I didn’t miss mention of it, but what was the sample size you used for the charts? Just 2015? Career totals for each pitcher? Marco Estrada’s an interesting case. His MLB-leading vertical drop on the changeup in 2015 was often mentioned as a reason for success but it really wasn’t that much better than 2014 (10.0 vs. 9.5 zMov). That can’t be the sole reason for his improvement in managing contact, and I’m wondering if this is where a recent change in pitch sequencing may have come in. Great article, btw. Definitely getting bookmarked for future reference!