Ye Gods of BABIP! An Ode to the Ground Ball

‘Twas the baseball season of ’99,
when Kevin Millwood was divine,
on many faces, we found egg
for some ranked him above Maddux, Greg

Alas! the BABIP gods were cruel,
Millwood’s rapid descent they did fuel,
A lesson in luck we all learned,
Never again would we be burned

Mediocre poetry aside, I remember the 2000 season vividly. It was before the days of mainstream analytics, when you had to calculate your fantasy roster by hand, or with Excel if you were really advanced. It was also a few months after the over-hyped Y2K disaster; somehow the world didn’t explode when ages-old operating systems thought it was 1900 instead of 2000.

Well, let me clarify, I remember very little of that season, except for the fact that I drafted Millwood instead of Maddux for the simple reason that Millwood had just come off a 2.68 ERA to Maddux’s 3.57 ERA. Millwood also had a tantalizing 0.996 WHIP, which clearly indicated he was a dominant force. This all came crashing down in 2000, and we were reminded that batted-ball luck is mostly just that — luck.

Let me pause from this bout of nostalgia and introduce the main thrust of this piece: peeling back the onion on BABIP a little, to reveal the gods (small g) above them. We’ll take a look at BABIP across a few vectors: trajectory (FB/GB/LD/PU), handedness (L/R), years and distance, as well as left/center/right. I don’t have data on whether a ball was hit soft/medium/hard, so we’ll have to make do without. I hope we’ll still be able to find some interesting nuggets.

We’ll look at some of the factors that affect BABIP and then pull out an arbitrary subset of pitchers who are consistent groundball BABIP over- or under-performers. (Check out Mariano Rivera’s groundball BABIP. You’ll need to scroll alllll the way to the bottom of the article if you want a sneak peek.)

We begin with a little housekeeping. All data in this article are sourced from PITCHf/x’s public data set and, as such, are subject to classification nuances that differ over time. Exhibit A shows the relative share of batted balls since 2008, with relatively stable groundball and pop-up (light blue) percentages. Interestingly, line drives have increased to almost 26 percent of all batted balls, pretty much fully at the expense of fly balls.

When I first saw these data, my instinct was to wonder whether there was a macro trend in baseball in which we were suddenly seeing a whole lot more line drives, perhaps a result of hitters going for a more contact-oriented approach at the expense of fly balls and power. I want to stress this is entirely not the case. It is almost certainly an effect of more fly balls being classified as line drives.

You can see from this chart that the increase in line drive percentage (and decrease in flyball percentage) is accompanied by sharp reductions in BABIP in both buckets, directly in line with the proportional increase/decrease. Line drives are being diluted with fly balls and thus are seeing the average BABIP reduced, while fly balls are losing their most line drive-like trajectories and thus reducing their effective BABIP.

What we see here is that when we combine all non-groundball balls in play, BABIP is very consistent year to year. If line drives indeed were up, we would see BABIP trending up, which it clearly is not. It is possible pop-ups are bleeding into fly balls, though the data set is too small to make any strong judgments.

Groundball BABIP

Now that we’ve gotten that out of the way, let’s look at ground balls. And as we know, not all ground balls are equals before the Gods of BABIP (big G). Ground balls are the most consistent from a trajectory classification standpoint, which makes for a good starting point. I computed a rough “angle” for the ball, representing how many degrees away from the left-field line a ball is hit, with zero being directly down the left-field line and 90 being exactly down the right-field line. The data are far from perfect (distance and angle will be skewed since hits and errors are charted where they were picked up) but overall work quite well.

For simplicity, I’ve ignored the platoon advantage, which we’ll touch upon later. The first histogram is for righties, the second for lefties. Each bar represents 1.5 degrees of the field, so the first bar is 0.0 to 1.5, the second 1.5 to 3.0, etc. The bars around the 45-degree mark represent a ground ball up the middle. I would surmise that any ground ball up the middle is being marked pretty much in line with the bag, which might explain the dip in BABIP right after. Overall, what we see are dips around 10 degrees (the third baseman), 30 degrees (shortstop), 60 degrees (second baseman) and 83 degrees (first baseman).

This graph shows how critical an infield defense is, as we see tremendous BABIP spikes in areas where coverage is weakest (i.e., ground balls up the middle). Now, as we will see in the next chart, lefties have a distinct advantage when hitting to the third-base side but are at a slight disadvantage to the first-base side (data exclude bunts).

The above chart begs the question of what is driving the improved BABIP to the opposite field. It appears almost counter-intuitive when you consider pull-side grounders should be hit harder than opposite-side grounders. Perhaps infield hits are skewing the results? Let’s look at where the most benefit is accrued to the left-handed batter:

The orange line represents left-handed hitters, while the blue line is right-handed hitters. The percentage is BABIP expressed in a percent, with the round number being the number of feet it traveled before it was fielded. We see a distinct advantage for righties beginning at 90 feet and continuing all the way to 190 feet away from the plate. This is essentially the prime “shift” zone, which would explain ground balls that travel 190 feet but end up as outs (other than the data being less than perfect). In fact, as the following table will show, the difference is largely based on balls hit toward right field. (For simplicity, I’ve split the field into five sections: left field, left-center, center field, right-center and right field.) Numbers represent BABIP differential to right-handed hitters. (Negative is worse for left-handers.)

We can see quite clearly that the biggest BABIP differences occur between 140 and 160 feet when the ball is hit to right-center or right field. There is a rather large BABIP boost for balls hit down the left-field line. However, as we will see in the next chart, this is by design, since the vast majority of ground balls are to the pull side.

This phenomenon explains the rationale behind shifting, as the vast majority of ground balls are hit to the pull side.

Groundball BABIP by Pitch Type and Trajectory

We see that four-seam fastballs (FF as classified by PITCHf/x) have a distinct advantage with respect to fly balls but a significant disadvantage with respect to ground balls, followed by two-seam fastballs, which do poorly across the board. As we saw in the beginning of this article, there is a lot of noise with respect to the line-drive/flyball classification, so I wouldn’t read too much into the line-drive BABIP except for maybe the curveballs, which are considerably higher. We’ll dig into fly balls and line drives in a future piece. For now, let’s investigate ground balls a little more and see if velocity has an impact.

We see a few interesting things here.

• Four-seam fastballs (FF) do not appear to benefit from velocity in any meaningful way. We do see a dip around 83-84 mph; however, those results likely are suffering from classification issues and are more likely than not mixed with change-ups. There is a spike in the 92-94 mph range, where groundball BABIP is the highest, and we are talking about a meaningful sample of roughly 50,000 ground balls in play, so there is a possibility that essentially slightly above-average fastball velocity is the sweet spot for batters to get groundball base hits. At the extreme top end, we do see a dip in BABIP, suggesting Aroldis Chapman-type velocity can have a BABIP impact (which would get lost in a simple linear model).
• Two-seam fastballs (FT) would suggest there is no mid-velo spike in BABIP but corroborate the top-end velo BABIP reduction abilities for fastballs.
• Change-ups (CH) and splitters (FS) perform very much the same and help to explain Johan Santana’s late-career BABIP suppression as well as helping to show how effective a well-placed splitter can be. Even a Koji Uehara-paced splitter barely cracking 80 mph will be almost useless if it’s hit on the ground.
• Sliders (SL) appear to gain effectiveness, in terms of BABIP suppression, the harder they are thrown, with a very narrow band between 79 and 85 mph of roughly .230, dropping to the .220 range from 86 mph and above. We’re talking one ground ball out of every 100, so it’s not a huge impact and probably difficult to measure for an individual pitcher, but there does appear to be a relationship.
• The curveballs (CU) is the only pitch type that appears to be less effective the harder it is thrown. My theory here is that the softer a curveball is thrown, the higher in the zone it will be, all else being equal. As we’ll see in the next chart, height of the pitch as it crosses home plate is critical for most pitches.

Pitch-Type GB BABIP by Height (feet) as the Pitch Crosses Home Plate

We see extremely linear relationships between groundball BABIP and the height of the pitch, especially in the off-speed pitches (CH, CU, SL, FS), and to a lesser extent FT and FF. Cutters do not appear to have such a relationship, which is interesting and probably worth investigating further.

Above is a chart that doesn’t really mean much on its own, but it looked pretty cool when I was messing around in Tableau. Essentially, it shows you what a specific rank in groundball BABIP will translate to in terms of actual BABIP. A rank of No. 20 will get you approximately a .300 groundball BABIP, whereas being around No. 200 will result in a .200 BABIP. This chart starts to make sense when we contrast it to the pitcher version.

It’s not a “yuge” difference, but there is more separation between the batters and the average, compared to the pitchers and the average, confirming the notion that batters have more control over BABIP than pitchers do.

Now, here comes the real question: Do any pitchers consistently rank in the top echelon of groundball BABIP? When I watch baseball, I definitely can tell when a groundball pitcher is generating weak contact, and it is most definitely a skill (in my opinion). In this case, we should see some pitchers who consistently outperform the average. In lieu of dumping a ton of data into a (yuge) table, I’m going to select a few interesting cases.

GB BABIP is Most Definitely a Pitcher Skill

Let’s compare Kershaw and Scherzer first. In every season since Kershaw’s rookie season, he has outperformed Scherzer significantly in the groundball BABIP metric, so much so that it probably explains a lot of the difference between the two from an ERA standpoint. In fact, Kershaw generates consistent surplus value from his ground balls, while Scherzer gets negative value.

One might argue this is due to variations in fielding (quite possible); however, I then would point to Melancon, who (single data point aside) puts up such consistent year-to-year groundball BABIP numbers that is screams legit skill to me. Dickey is easy to explain with the knuckleball, but we also see pitchers who are mediocre — such as Colon, Harang and Papelbon — post consistently poor groundball BABIP numbers that are at best average (+/- outliers). Ziegler is dominant at preventing hits from his ground balls.

Rivera, with Derek Jeter as his shortstop, posted ridiculous groundball BABIP numbers. Perhaps a signature of greatness is the ability to completely devalue the ground ball, providing more consistent dominance.

Concluding Thoughts

We definitely need to delve into the why a little more, with respect to pitchers being able to control groundball BABIP. My gut says it may be an indicator of command, though I have little in the way of expressing that numerically.