A Look Into the Mist: Pitcher Defense

When considering how to evaluate a pitcher, there are many philosophical roads that can be traveled. You can choose to look at total runs allowed, you can choose to use peripherals like strikeouts and HR/9, or you can go more in-depth with data that reveals more about pitch strategy and contact quality. Whatever strategy one chooses for said approach, it is vital that a pitcher's defensive contributions not be forgotten.

One simple fact is that pitchers are part of a defense, and therefore we know at least a part of what is happening behind a pitcher after the ball is put into play. Defensive efforts should be intertwined with all of the roads of pitcher analysis. But really, what can we do with pitcher defense? What exactly is there to add to the already complex equation to help us understand pitcher defense? Well, what about a metric that adds a defensive component to xFIP or a similar metric? That part is just food for thought for readers, but I do want to get into the impact pitcher defense has on performance.

There are many different components that go into pitcher defense. The first part is the fact that pitchers come off the mound and actually field balls. They aren't just zombies sitting out on the mound after they throw a pitch. However, there are other parts to their defense, too. Just as with catchers, pitchers have a huge responsibility in shutting down a running game. Their times to the plate and pick off times are a vital component of understanding the importance of pitcher defense. After all, being able to limit the movement of runners on the bases is a key component in run prevention. The further you keep a base runner from home plate, the less likely he is to be able to score, theoretically. The first thing I want to do here is eliminate pitcher defense entirely. I want to do this to show what results would be like if every pitcher truly was only impacted by the other eight defenders behind him.

Before I blab on for too long, here is a table showing an adjusted version of RA/9 that removes the positive and negative pitcher defense runs from a pitcher's performance (I'm using Defensive Runs Saved as my defensive metric, and to save space, these are the highlights):

Top 15:

Jered Weaver	2.80
Clayton Kershaw	2.87
Cliff Lee	2.92
Justin Verlander	3.02
Chris Sale	3.05
Cole Hamels	3.07
Matt Cain	3.20
Roy Halladay	3.21
Johnny Cueto	3.26
Jeremy Hellickson	3.30
Gio Gonzalez	3.30
David Price	3.31
Josh Johnson	3.33
Ryan Vogelsong	3.41
Jordan Zimmermann	3.43

Bottom 15:

Jason Marquis	5.42
Jeff Francis	5.50
Ubaldo Jimenez	5.52
Livan Hernandez	5.54
Tyler Chatwood	5.54
Luke Hochevar	5.61
Francisco Liriano	5.63
Brian Duensing	5.63
Chris Volstad	5.65
Tommy Hunter	5.78
Jake Arrieta	5.85
Brad Penny	5.85
Jordan Lyles	6.01
Roberto Hernandez	6.30
Nick Blackburn	6.41

As I mentioned, this number reflects what the pitcher's RA/9 would be if the impact of their own personal defense was completely removed, assuming DRS is accurate in accounting for the number of runs a pitcher prevented or allowed in the field. This list doesn't mean anything without context, so here's another leader board, this time featuring what the quantifiable difference is between RA/9 and DRS-removed RA/9):

Top 15:

Jake Westbrook	0.48
Mark Buehrle	0.42
R.A. Dickey	0.33
Jhoulys Chacin	0.31
Johnny Cueto	0.27
Mike Leake	0.26
Joe Saunders	0.26
Henderson Alvarez	0.25
Ricky Romero	0.24
Jeff Francis	0.24
John Danks	0.24
Zack Greinke	0.23
Clayton Kershaw	0.21
John Lannan	0.21
Tommy Hunter	0.21

Bottom 15:

Zach Britton	-0.17
Ervin Santana	-0.18
Ricky Nolasco	-0.18
Brett Myers	-0.19
Brandon Morrow	-0.21
Kevin Millwood	-0.21
Josh Beckett	-0.22
Matt Garza	-0.24
Mike Pelfrey	-0.25
Ted Lilly	-0.26
Freddy Garcia	-0.28
Tommy Hanson	-0.30
Derek Lowe	-0.30
Edinson Volquez	-0.31
Philip Humber	-0.34

As you can imagine, some of these results are staggering. Jake Westbrook, with as mediocre as his RA/9 numbers are, has shaved off nearly a half a run per nine innings due to his outstanding defense over the last two years. This is in stark contrast to pitchers like Edinson Volquez and Phillip Humber who are worsening their own efforts by not fielding their positions well.

It should be noted that there are all kinds of pitchers here, too. High strikeouts guys, low strikeout guys, high GB% guys, low GB% guys, high LOB% guys, and low LOB% guys. What I conclude from this is that fielding a position probably has more to do with how a pitcher comes off the mound or moves to the plate than it does in the actual style he pitches with.

For instance, Johnny Cueto is known to be quick to the plate and gets quite a few ground balls (50+% in years past). R.A. Dickey, despite being a knuckle ball pitcher, has only allowed 20 stolen bases in 29 attempts dating back to the beginning of 2009. This 31% rate is actually three percent better than the league average over that same time span. Inversely, we can look at a guy like Brandon Morrow. Morrow, over the past two seasons, has allowed 21 stolen bases (more than Dickey) in 27 attempts. This 22.2% CS rate is well below the league average over that same time span.

Some of this is going to depend greatly on who the pitchers have faced and the catchers that have been behind the plate (along with a multitude of other variables), but the results tend to be pretty consistent.

The next set of questions that came to mind in my head is, "Well, are pitchers who are good at defense any better at preventing runs? Are these pitchers typically ground ball pitchers? Does this skill account for the difference between their ERA and a statistic like xFIP?"

The answer to all three of those questions is a resounding, "Answer not clear." While this is a great concept to understand (pitcher defense itself), it still doesn't show much of anything in accounting for trends and differences. There are too many individual variables in play here to truly gauge the impact of defense. Here is a small list of correlations I ran and the R values that were produced from the relationships:

DRS/IP vs. GB%: .069 (measure of DRS per inning vs. ability to produce ground balls)

DRS/IP vs. LOB%: .187 (measure of DRS per inning vs. ability to strand runners)

DRS/IP vs. BABIP: .101 (measure of DRS per inning vs. ability to prevent balls successfully going for hits)

DRS/IP vs. ERA-xFIP: .219 (measure of DRS per inning vs. difference between raw ERA and xFIP)

While none of these values are really conclusive, it definitely is encouraging to see that the relationship between defensive efficiency and skills such as LOB% and the difference between ERA and xFIP is higher than in other relationships. If nothing else, DRS ends up giving us another tool to use when considering why a pitcher's fielding-independent pitching doesn't quite match up with his ERA. The ability to defend one's position while on the mound is a big part of pitching, and it's also a repeatable skill, which plays big into analyzing his abilities as a player. After all, there are some things pitchers can do to help their success once the ball is put in play against them, and this is one of them. Therefore, it's time to take a crack at developing a new metric: DRS% Score (name change likely, suggestions recommended).

DRS% Score

Essentially, the metric is this: take the number of defensive runs saved or lost and divide it by RA+DRS. What this tells you is the volume of defensive runs saved or lost by the influence of the pitcher relative to the number of runs that would be given up if the pitcher was not there. Then, divide that number by the league average. After all is said and done, the equation looks like this:

(DRS/(DRS+RA)) - (lgDRS/(lgDRS + lgRA))

Under this system, the league average DRS% Score will be zero. Poor scores will be large negative numbers, and good scores will be large positive numbers. Using this, these are the pitchers that had the largest positive and negative influences on their performance through their defense over the last two years (who pitched at least 200 innings):

Top 15:

Mark Buehrle	0.089
Jake Westbrook	0.085
R.A. Dickey	0.083
Johnny Cueto	0.075
Clayton Kershaw	0.068
Jhoulys Chacin	0.062
Justin Verlander	0.055
Joe Saunders	0.051
Zack Greinke	0.050
Mike Leake	0.049
Hiroki Kuroda	0.046
Ricky Romero	0.044
Henderson Alvarez	0.042
John Danks	0.040
John Lannan	0.039

Bottom 15:

Ricky Nolasco	-0.045
Ervin Santana	-0.047
Brett Myers	-0.049
Josh Johnson	-0.049
Kevin Millwood	-0.053
Brandon Morrow	-0.056
Mike Pelfrey	-0.061
Derek Lowe	-0.065
Josh Beckett	-0.068
Matt Garza	-0.068
Edinson Volquez	-0.073
Freddy Garcia	-0.074
Ted Lilly	-0.074
Tommy Hanson	-0.078
Philip Humber	-0.081

I like the way these numbers came out, because this metric helps create context around just how much the pitcher is contributing to his own cause. If anyone has any ideas for tweaking this, please contact me.

Conclusion

There is certainly work to be done here as far as pitcher defense is concerned. With results as extreme as Mark Buehrle's, it's obvious that being a strong defender carries value (sometimes very significant value). When it comes to projecting performance, I'm a big fan of players that have repeatable skills, because I know a lot more about what I'm going to be getting.

This doesn't make plus defenders at the pitcher position inherently more valuable than other pitchers, but it provides another tool for accurately gauging their value. I stand in a funky place between using fielding independent pitching and RA/9 numbers to value pitchers, and a big part of it all is the fact that I think it's possible to accurately analyze pitcher defense. The next time you look at a pitcher's Fangraphs page and see that his ERA was much better than his fielding-independent metrics, consider his skills as a defender (as well as his LOB% and a multitude of other things). After all, in the complex world of pitching, it's good to have as many repeatable skills as possible. Lady luck is a fickle beast, so removing her from the equation as much as possible is ideal.

---

Big thanks to baseball-reference.com and fangraphs.com for making it super easy to go out and get the data for this project.