The Pirates and building ground ball pitchers

Voluminous quantities of ink/ink’s electronic analogue have been spilled discussing the extreme ground-balling tendencies of the Pirates’ pitching staffs of recent years. Last fall, Tony Blengino of FanGraphs detailed the unprecedented extent to which Pirates pitchers generate ground balls (and weak ground balls, at that). This team-wide philosophy of inducing ground balls into a heavily shifted infield defense has, according to analysis performed by Bucs Dugout’s own David Manel, possibly been worth as much as fifteen wins over the last two seasons. Or roughly one Andrew McCutchen.

The Pirates have targeted and acquired pitchers like Francisco Liriano and A.J. Burnett; they’ve also developed their own (Jared Hughes, Justin Wilson). Many pitchers have increased their ground ball rates upon arriving in Pittsburgh, indicating that the team doesn’t view ground ball generation as a wholly static quality, but rather something that can at least partially be taught.

Consider the following pitchers and the changes in ground ball percentage they experienced during their respective first seasons in Pittsburgh:

Most statistical projection systems (ZiPS, Steamer, PECOTA) essentially view players as probability distributions, with certain levels of performance weighted as more likely than others based on a player’s track record. What makes pitchers particularly hard to predict is their ability to change aspects of their approach (arsenal, location) so as to become substantively different players. While the guys above were, generally, ground ball pitchers prior to being acquired by the Pirates, they subsequently began generating even more grounders. The question is: how? What makes a ground ball pitcher?  Or, more specifically, what did the Pirates see in these guys that suggested they could become ground-ball machines?

Received baseball wisdom tells us that working down in the zone leads to more ground balls, and we know that Pirates’ pitching coach Ray Searage is a proponent of the two-seam fastball, a ground ball inducing pitch. But if it were as simple as telling everyone ‘hey, throw more two-seamers and keep the ball down,’ there wouldn’t be any need to target specific players, as any pitcher could theoretically become a grass-killing dynamo. It seems as though, to borrow a phrase from developmental psychology, ground ball tendencies are part nature, part nurture.

To determine how some pitchers generate ground balls than others, one could turn to this excellent article for discussion of ‘heavy’ pitches from a physics perspective. Or read former MLB pitcher Zach Day’s article on the relationship between spin, whiff rates, and ground ball rates. Both contain a ton of useful information (at least, I assume the physics one does—maybe I should’ve paid more attention in high school?).

What I’m going to do, though, is try to determine what the available information tells us. In this particular article, I’m going to look at the ‘nature’ side of things, specifically in relation to Searage’s pitch of choice, the two-seam fastball.

Using Baseball Savant, I downloaded every two-seam fastball put in play during 2013 and 2014. Significantly, Baseball Savant classifies two-seam fastballs and sinkers separately (although they’re similar pitches), which means, for example, that Jared Hughes and A.J. Burnett won’t be in our sample.

Not all two-seam fastballs are created equal

Why does quantifying a pitcher’s stuff matter? As Day observed, spin, velocity, vertical break, and pitch location all contribute to a pitcher’s ground ball-inducing ability. Check out these graphs of the relationship between pitchers’ two-seam fastballs’ average spin, velocity, and vertical break and the resultant ground ball percentage they generate:

All of these relationships are significant at a 99% confidence threshold, but the correlation is much stronger for spin and vertical break (as you can probably see). So to induce more ground balls, you generally want your two-seam fastball to have low spin, high velocity, and high break.

If you prefer looking at the individual pitch level, the following chart plots the in-play two-seam fastballs by spin rate and break, color-coding them by batted ball type:

Looking at this graph, a few things become clear: first, extremely low-spin two-seam fastballs necessarily are going to have relatively high vertical break almost as a rule (the laws of physics being what they are). Second, it’s pretty tough to hit low-spin, high-break pitches in the air.

Determining a pitcher’s ‘stuff’

We’ve established that there are different kinds of two-seam fastballs—but are spin rates and break length innate attributes of a pitcher’s fastball? After all, if they vary substantially from year to year, we may just be capturing noise.

Here are charts of the year-on-year stability pitchers experience for all three measures:

Plotting the year-on-year stability in velocity reveals pretty much what you’d expect—very few guys experience major year-to-year changes in two-seam fastball velocity. You can see one outlier on the graph—I’ll discuss him below.

Spin, however, is also very stable—very few pitchers add or lose significant spin between years.

Ditto for break length—high-break guys continue to be high-break guys, and low-break guys continue to be low-break.

You’re probably sick of scatter plots by now, but hopefully they’ve made their point: pitchers’ two-seam stuff is almost entirely constant season-to-season. We know what type of velocity, spin, and movement a given pitcher’s two-seam fastball will have next year. This is a repeatable skill.

The darker dot on each graph is Charlie Morton—I’ll discuss him again later, but for now just note that he throws one of the lowest-spin, highest-break two-seamers in the game, despite his mediocre velocity.

The big outlier in the velocity graph is Orioles’ closer Zach Britton, who increased his two-seam fastball from 91.6 to 95.1 MPH between 2013 and 2014. Britton’s spin rate also experienced one of the largest increases, from 1,982 to 2,145 RPM, and his vertical break correspondingly decreased from 6.6 to 6.0 inches. There’s a simple explanation for this, however: Britton was converted from a (mediocre) starter into a (decidedly better than mediocre) reliever between 2013 and 2014. Already a pitcher with considerable ground ball tendencies, Britton’s ground ball rate on two-seam fastballs increased from 66.7% to an utterly ridiculous 77.4%, which might seem bit surprising given his increased spin and decreased break.

So now we know that certain types of two-seam fastballs induce more ground balls than others, and we know what sort of each fastball each pitcher possesses. The next missing piece is location.

Location matters

Stuff isn’t the only thing that matters when generating ground balls. Managers and pitching coaches have long preached that keeping the ball down will lead to more ground balls, and as we can see from the following graphics that appears to be true. Use the zone map (again courtesy of Baseball Savant—it’s amazing all the information they provide) to interpret the ground ball rates by pitch location:

Zones 7-9 and 13-14—those at the bottom of the zone and outside the bottom half of the zone—have by far the highest ground ball rates. Zones 1-3—those at the top of the zone—have the lowest ground ball rates.

If you prefer a more intuitive visualization, the following chart plots batted ball outcomes by horizontal and vertical pitch location, revealing that it’s not *just* pitching at the bottom of the zone that helps generate ground balls—it’s also staying away from the middle of the plate. In general, getting hitters to swing at pitches outside the strike zone is a great way to induce weak ground balls.

The orange blob (I’m using the technical term here) is almost completely within the green one.

So what?

None of this, thus far, is particularly revolutionary on its own. Maybe you didn’t know about the relationships between spin, break, and ground ball rate. I didn’t, before I read Day’s article and looked into the data on my own.

But what this allows us to do is develop a model to predict whether a particular ball in play will be a ground ball given its location, spin, velocity, and break. We can also build a similar model that ignores location, isolating only the ‘stuff’ components that drive high ground ball rates. I used logistic regressions to accomplish both of these tasks (gory details available, should any particularly masochistic readers be interested), then added up the predicted values to reconstruct an ‘expected’ ground ball rate on two-seam fastballs for every pitcher—one overall value, then a ‘stuff’ value.

I’m not going to deal with the zone-based model in this article, but the ‘stuff’ model can deliver some interesting insights. Let’s look, for example, at the pitchers whose two-seam fastballs should’ve, purely based on raw stuff, generated the most ground balls in 2013:

The only guy on the list whose two-seam fastball doesn’t generate at least 50% ground balls is Chris Sale. More notably, the names that pop up here include Charlie Morton, the previously mentioned Zach Britton, and the Pirates’ own Gerrit Cole.

If we compile the same table for 2014, there’s a predictably large amount of overlap:

Morton is still near the top—the low spin and significant vertical break on his fastball allow him to generate a ton of ground balls. If the Pirates’ pitching staff is an unprecedented experiment, Morton is their Frankenstein. Eternal Pirates nemesis Lance Lynn and the recently well-traveled Jeff Samardzija are the other starters to appear on both lists.

What’s the utility of this exercise? For one, it could help identify pitchers who have the raw stuff to induce lots of ground balls, even if they’re not doing so presently. It could also lead us to a better understanding of how different pitchers manage contact. I wouldn’t have guessed, prior to performing this analysis, that the vertical break on a pitcher’s two-seam fastball was of paramount importance in generating ground balls.

Three pitchers, in particular, jumped out to me as interesting cases (and potential Pirates acquisitions—who knows?). Javier Lopez, T.J. McFarland, and Aaron Loup aren’t household names, but they’re quite extraordinary in that they have some of the ‘heaviest’ stuff in the game. This is even more fascinating because none of the three have particularly impressive velocity, and none of them have particularly low spin—they generate ground balls largely through tremendous vertical break.

Javier Lopez is a 37-year-old lefty specialist currently with the Giants. In 2013, an astounding 80.3 percent of the contact off of Lopez’s two-seam fastball was on the ground, with only a slight regression in 2014. Possessing only an 85 MPH fastball, Lopez has nonetheless comfortably outperformed his FIP each of the last six years by maintaining the highest vertical break in the game.

Lefty swingman T.J. McFarland hails from another organization that seems to value contact management—the Orioles. As with Lopez, McFarland’s calling card is a high-break fastball, which is surprising given that his two-seamer has well above average spin. McFarland is capable of both starting and relieving, and profiles as a pretty similar pitcher to Jeff Locke. I wouldn’t be at all surprised if his name was discussed during the Travis Snider deal.

Toronto reliever Aaron Loup (also a lefty—maybe I need to investigate handedness further) has a very unique two-seam fastball—it’s among the league leaders in both spin and break, which, according to the laws of physics, should be unlikely. Perhaps subverting gravity is the new market inefficiency.

We still don’t know exactly how to quantify pitchers’ ability to manage contact. Maybe we never will, even with the new Hit(fx) data that’s starting to become publicly available. But it seems clear that looking at measures like spin and break can lead us toward a more complete, more nuanced understanding of what makes each pitcher tick.