FanGraphs Spotlight: Pitch Type Splits

This month at FanGraphs, we’ve been highlighting a number of site features and showing you how we use them. The goal is to make your visit to the website more useful, and to get the most out of the features we’ve added over the years. Today, I’m going to walk you through the data we keep on individual pitch types: how you can look at what’s in your favorite pitcher’s arsenal. I’ll also show you how to do this for every member of your favorite team’s pitching staff, as well as for all the pitchers in the majors.

One quick note here before we get started: All of this data can get chopped up any way you’d like using our custom leaderboards, which Dan Szymborski explained earlier this month. If you’re already a power user, this is just more data to pour into the soup. I’ll be looking at our standard-issue pages today for ease of use, but please feel free to mix and match these site tutorials in any way you prefer.

One of the greatest advancements in baseball data collection came in 2008. That year, Major League Baseball started publishing data produced by a system called Pitchf/x. Since then, we have location, speed, and movement data for every pitch thrown in the majors, give or take a few one-off stadiums that at the time didn’t have the correct camera system installed.

Before this data existed, we described pitches qualitatively. One curveball might look “sharp” while another looked “lazy,” but that’s as far as analysis could go without numbers to back it up. In an instant, that all changed. Ever since then, we’ve been able to quantify these details, and you can look up every one of them on the pages of FanGraphs.

Let’s use a great pitcher who happened to debut in 2008 as an example. Clayton Kershaw’s three pitches are all famous – and they’re all measured for posterity, down to the fraction of an inch. To look at this data, first head to Kershaw’s player page. There, under the “Splits” tab, you’ll see two possible types of pitch splits:

The two types of pitch splits, Pitch Info and Statcast, refer to who is classifying the data. The two are slightly different, most frequently when it comes to deciding what pitch type fastballs and hard breaking balls belong to. For the duration of today’s article, I’m going to stick to Statcast, which refers to the pitch types assigned by Major League Baseball, first using Pitchf/x and then its later successor systems. The two have broadly similar conclusions – almost every pitch gets classified the same way by both systems – but if you’re an advanced user and want to check out the differences between the two, I thought I might as well explain what separates them.

OK, so you’ve selected a pitch type split. You’ll see a huge pile of data for every year where the pitcher you’ve selected threw at least one pitch of that type. The first grouping on that page is some basic data:

The most relevant data points in this grouping are number of pitches thrown (which you’ll see in every tab) and velocity. You can see the slowest, fastest, and average speeds for each pitch. Kershaw, as you can see, varies his slider quite a bit, with 10 mph gaps in most years. I also like looking at home runs allowed; it’s a good way of spotting whether a particular pitch suddenly became too hittable, though it’s best to remember that it’s a very noisy statistic.

The next tab is actually my least favorite:

There’s some good data here, but it misses some key context. Each of these columns looks at how batters fared on plate appearances that ended with a slider, ignoring all other sliders. You can see that Kershaw’s slider has generally allowed a low BABIP, and that it’s ended a lot of plate appearances with strikeouts. But I’m also interested in how the pitches that didn’t end plate appearances fared, so let’s press on.

The next tab is a great one:

Now we’re talking. Wondering if a pitch induces a lot of grounders when an opponents puts it in play? You can find it here. Kershaw’s slider does indeed get a lot of grounders, and a lot of infield fly balls too (IFFB% is the percentage of fly balls that don’t leave the infield). The top end for grounders is much higher than this – Logan Webb’s sinker gets grounders at a 65% clip – but you can see at a glance that Kershaw’s slider has generated plenty of the best types of contact, grounders and pop ups.

The three columns on the far right are why we consider 2008 to be the start of the pitch tracking era. Thanks to the new cameras and the liberal use of physics equations, we have access to how much a pitch “moved” on its path to home plate. The exact specifics of those calculations are beyond the scope of this article, but here’s a rough primer.

The first column, xMov, measures how far the ball moved horizontally compared to the path you’d predict if it traveled in a straight line from the pitcher’s release to home plate, in inches. Negative numbers mean that the ball is traveling toward a right-handed hitter – they’re given from the perspective of a catcher, with negative being to the catcher’s left.

The next column, zMov, measures how much up or down a pitch moved, relative to a spinless path home. Of note, zMov excludes gravity. In other words, Kershaw’s sliders all fell from his release point on their path home, but in every year except 2010, the spin he imparted on the ball made them fall by less than they would have if they’d merely been dragged down by gravity.

The last column, Mov, simply measures the total movement of the pitch in a straight line. If you remember your middle school math, it’s just the result of constructing a right triangle with xMov and zMov as two of the sides and measuring the hypotenuse. We can see, in this example, that Kershaw’s slider gained a ton of vertical movement in 2017 and 2018, but that he’s dialed that back down in recent years.

Before we move on to the last tab, let’s do a quick comparison. Kershaw’s slider breaks gloveside and rides slightly, resisting gravity’s pull. His curveball, on the other hand, dives:

Combine that with the fact that he throws the curveball nearly 10 miles an hour slower, and you can understand why batters end up taking such bad swings at these two pitches. They’re both breaking balls, and they both break gloveside, but their movement diverges wildly on the vertical axis. Sit on a slider and get a curveball, and you’re probably going to look foolish.

OK, back to the slider. The last tab, plate discipline and value, is perhaps the most useful. It looks at every single pitch thrown and tabulates the results of hitters’ swing decisions, completely ignoring what happens after the moment the bat meets (or doesn’t meet) the ball:

Let’s go through these in order. First, you have O-Swing% and Z-Swing%, how often hitters swing at pitches outside and inside the strike zone, respectively. Swing% is straightforward. O-Contact% and Z-Contact% measure how often swings connect with pitches in those zones. For a point of reference, Kershaw’s fastball has a career O-Contact% of 73.6%, so you can see that his slider is missing *far* more bats when hitters chase. The fastball has a Z-Contact% of 87.4%, so the slider is missing more bats even in the zone. That’s why pitchers throw sliders, all quantified for you.

The most important columns are the last few. SwStr% is one of the best statistics there is for predicting a pitch’s effectiveness. Called strikes aren’t sticky from one count to the next; hitters might let a pitch go at 1-0 that they’d clearly swing at if the count were 1-2. Batted ball results are famously noisy. But throwing one past a guy? There’s not a lot of room for interpretation; you just threw one past him. Kershaw consistently finishes with around 20% of his sliders turning into swinging strikes. To stick with the fastball comparison, his career mark there is 7.3%. Sliders are hard to hit!

Finally, we’ve got pitch value, and pitch value per 100 pitches. That measures what actually happens as a result of each pitch. It takes more than just final results into account, too; the value of moving from a 2-0 to 2-1 count is significant, as is the lost value from moving from 2-0 to 3-0. You can even think of this in terms of its effect on ERA with a little massaging. Assume around 150 pitches per game, multiply the run value per 100 pitches by 1.5, and you can think of how much, say, Kershaw’s slider affects his ERA, though in reality the cause and effect aren’t nearly so clear. But as a first order approximation, it’s just fine. Kershaw’s slider is roughly two runs above average per 100 pitches, so per 150 it’s three runs above average; in other words, it’s really good.

One way to use these pitch splits is to gawk at the greats of our era. You can also use them to look for changes in pitch shape over time. Take Kershaw’s former Dodger teammate Blake Treinen, for example. His slider went through a significant transformation in a single offseason:

That’s when the Dodgers were teaching everyone sweepers. If you were evaluating his results from one season to the next, it would be useful to know that his slider gained half a foot of horizontal movement. Without pitch splits and movement data, that would be nearly impossible to tease out.

If you’re interested in generalizing these values and looking across the entire league, I highly recommend creating your own custom leaderboards, as Dan laid out. But we do keep some of the data on our default leaderboards anyway. Let’s say you’re interested in who has the sweepiest slider in baseball. Head to our 2023 pitcher leaderboards, then select Pitch-Level Data and H-Movement:

Hey, Mitch Keller fan club rise up! You can find a lot of fun data points by hopping around those leaderboards. Justin Steele gets the most break among lefties. Blake Snell and Justin Verlander throw the fastballs with the most vertical movement among qualified starters; lower the innings threshold, and Félix Bautista and Nick Pivetta are the top duo. There’s a world of data out there, and since 2008, pitch shapes are part of that world. You can get as deep into this data as you’d like; some analysts comb through this data and highlight changes, breakouts, and surprising shapes as their full-time job. But even if you just want to dabble in this new(-ish) area of baseball quantification, FanGraphs has you covered.