### Introduction

As a Phillies fan, I’m obviously excited that Jake Arrieta is now on the Phillies pitching staff. They were desperately looking for a veteran to solidify the starting rotation and it appears that Arrieta will fit the bill. He won the NL Cy Young award in 2015, but his performance has declined the last two seasons. He spent much of the hot stove season unsigned as a free agent and the story was that many teams were reluctant to give him a long-term contract due to his decreased fastball velocity. This discussion motivated some questions to explore:

- What pitches does Arrieta throw?
- Do we see a decrease in speed in his pitches over the last three seasons?
- What about the movement of these pitches?
- How has Arrieta’s pitching effectiveness changed from 2015 to 2017?

As a side point, I’ll be trying out the `cowplot`

package that provides some add-ons to the `ggplot2`

graphing package. (You’ll see the graphs below plotted using the default theme provided by this package.) . I’m using Statcast data for the 2015 through 2017 seasons.

### What Pitches Does Jake Throw?

First let’s explore Arrieta’s pitch types. He throws five pitches, a changeup (CH), a curve ball (CU), a four-seam fastball (FF), a sinker (SI), and a slider (SL), but as the graph indicates, the distribution of these pitches has changed from 2015 to 2017. In 2015, he appeared to throw a high number of sinkers and sliders and a good fraction of four-seamers. In contrast, in 2017, he seemed to rely on sinkers and fewer sliders and four-seamers.

### Release Speeds of These Pitches

The concern about Arrieta was on the decrease in his fastball speed. Below I display boxplots of the release speeds for each of the pitches, comparing the three seasons. This demonstrates that the average speed in his four-seamer has dropped a few mph from 2015 to 2017. But this decrease in speed carries over to all of his pitches. Even his curve ball tended to be thrown at a smaller speed in 2017. (By the way, I illustrate the `cowplot`

package’s capability to insert a graphic on a `ggplot2`

figure.)

### Horizontal and Vertical Breaks

Since movement is an important component of a pitch, let’s explore the horizontal and vertical movements of these pitches for the three seasons. I notice significant changes in the horizontal movement of his four-seamer and curve ball in 2017. Also the movement of his other pitches also shows changes to larger pfx_x values. The obvious conclusion is that Arrieta is not throwing these pitches in 2017 as he was in the 2015 season. (Here I’m using the function in the `cowplot`

package that allows one to combine ggplot2 objects in a single graph.)

### Swing and Miss Rates

One measure of pitching effectiveness is the fraction of swung pitches that are missed by the batter. Below I have displayed these miss rates for all pitch types and all seasons. Arrieta’ miss rate for a changeup increased in 2017, but he had lower miss rates for his sinkers and four-seamers. (Of course, as one would expect, miss rates for off-speed pitches are greater than the miss rates for fastballs.)

### Exit Velocities

Another measure of pitching effective is the quality of the balls that come off of the bat. If the batter is not getting good wood on the ball (say, if he was fooled by the pitch), then that would result in a lower exit velocity and a less-than-optimal launch angle. (In other words, the batter may be more likely to hit a ground ball or a pop up.) First I show boxplots of the exit velocities for each pitch type and season. The basic message I get from this graph is that the exit velocities on balls in play haven’t changed much from 2015 to 2017.

### Launch Angles

What about launch angles? Below I display boxplots of launch angles for all pitch types and seasons. Here’s an interesting observation — batters have appeared to hit balls at larger launch angles — this is most clear for sliders, curve balls, and change ups.

### Home Runs Allowed?

Since launch angles tend to increase from 2015 to 2017, I thought I’d next look at home run rates (on balls in-play) on different pitches. One has to be cautious in showing these rates due to the small sample sizes, so I’ve labelled each point by the home run count. What is most striking is the 14 home runs that Arrieta allowed on sinkers in 2017 (contrasted with only 2 in the 2015 season).

### Summing Up

How has Jake Arrieta’s pitching changed from 2015 to 2017? We see from this exploration that there have been many changes. He is using a different distribution of pitches (more sinkers, fewer sliders and four-seamers). Moreover, the characteristics of these pitches have changed — the release speeds and movements are different. Now these observations don’t necessarily imply that Arrieta was less successful in the 2017 season. But we see changes from 2015 to 2017 in the miss rates and in the launch angles (and home run rates) of some of the pitch types which may be of concern.

The Phillies have an analytics group and I would assume they would know about these characteristics before they made the decision to sign Arrieta to a three-year contract. But clearly Arrieta will be a positive role model for the young pitchers on the Phillies staff. Given the changes that I’ve observed above, I’m very interested in exploring his pitch data as the 2018 season unfolds.

### R Code (added 3/20/18)

After I posted this, several people were interested in seeing the R script — I just posted it here on my GitHubGist site. The interested reader is encouraged to try out and improve on my graphics.

Is this work on GitHub?

I just added a link to the R script in my post.

Is this analysis available on GitHub?