Pitch Type xwOBA on Contact (xwOBAcon)

In 2018, and again earlier this year, I reviewed how different pitch types perform by various measures including swinging strike rate (SwStr%), ground ball rate (GB%), and isolated power (ISO). In the last couple of years I have tried to emphasize heavily the importance of evaluating a pitcher on his component parts — namely, each of his unique pitches, all of which behave differently and can bring resolution to some of pitching’s more enigmatic questions and issues.

If you clicked through those links in the first sentence, you saw how breaking balls and offspeed pitches outperform fastballs by virtually every metric. With the advent of Statcast, we can not only validate my prior work, which relied on PITCHf/x data, but also dig more deeply into how each pitch type behaves according to newfangled Statcast data — namely, how each pitch performs exclusively on balls in play.

This is something I pursued preliminarily using the PITCHf/x data, by measure of ISO, but it doesn’t fully capture total production or damage allowed. Having written about Zack Wheeler the other day and in discussing how the performance of his pitches have ebbed and flowed from 2018 to 2019, I was curious to dig into pitch-specific expected weighted on-base average (xwOBA) on contact (xwOBAcon).

Here’s how every pitch type compares by xwOBA allowed. Keep in mind, xwOBA captures “deserved” total value through not only balls in play but also strikeouts and walks. Year in and year out, fastballs fare worse than the league average, whereas breaking balls and offspeed pitches perform better than average, all to varying degrees.

Pitch Type xwOBA
Pitch Type 2015 2016 2017 2018 2019
Sinker .344 .351 .355 .343 .363
Four-Seamer .338 .349 .354 .347 .356
Two-Seamer .342 .354 .366 .351 .353
Average .308 .315 .321 .311 .318
Cutter .308 .311 .305 .309 .308
Change-up .284 .287 .301 .284 .289
Knuckle Curve .235 .241 .258 .260 .273
Curveball .240 .248 .260 .254 .266
Slider .251 .261 .261 .254 .268
Splitter .263 .273 .268 .252 .253
Sorted descending by 2019 xwOBA. Click headers to sort.

Generally the same can be said for actual wOBA, which, given the assumptions that underpin xwOBA, it should. At the league level, xwOBA approximates wOBA, much the same way FIP or xFIP equals ERA when rolled up to the league level. Thus, I will use xwOBA and xwOBAcon exclusively henceforth.

Pitch Type wOBA
Pitch Type 2015 2016 2017 2018 2019
Sinker .342 .351 .353 .344 .366
Four-Seamer .341 .350 .351 .346 .357
Two-Seamer .347 .351 .363 .352 .355
Average .312 .318 .322 .315 .321
Cutter .309 .312 .305 .312 .300
Change-up .291 .293 .301 .292 .289
Knuckle Curve .245 .250 .264 .271 .285
Curveball .250 .255 .266 .270 .276
Slider .262 .267 .272 .263 .276
Splitter .269 .280 .274 .252 .260
Sorted descending by 2019 wOBA. Click headers to sort.

What occurred to me in looking at Wheeler is the efficacy of breaking balls and offspeed pitches in terms of generating swinging strikes (and, thus, strikeouts) gives those pitches an unfair edge. Maybe “unfair” isn’t the correct word — their superior performance is well-deserved — but perhaps those whiffs mask underlying contact quality that is not necessarily superior to fastballs.

Here is each pitch type’s xwOBAcon by year:

Pitch Type xwOBAcon
Pitch Type 2015 2016 2017 2018 2019
Four-Seamer .372 .385 .395 .388 .409
Sinker .357 .368 .369 .359 .377
Knuckle Curve .338 .342 .354 .368 .374
Average .352 .364 .373 .362 .374
Two-Seamer .352 .365 .377 .360 .370
Cutter .345 .353 .359 .359 .356
Slider .332 .349 .354 .343 .355
Curveball .334 .345 .352 .345 .354
Change-up .332 .339 .352 .330 .338
Splitter .334 .343 .357 .339 .324
Sorted descending by 2019 xwOBAcon. Click headers to sort.

This table isn’t dramatically different, but you can see the playing field is much more level when isolating performance on balls in play. Whereas xwOBA (and wOBA) were ordered almost identically by pitch type year to year, xwOBAcon bounces around, with, for example, splitters greatly outperforming sliders in one year and underperforming them in another. You can easily substitute any other pair of pitch types into that previous sentence and have it remain valid.

While a couple of observations appear to be truisms (splitters are best, change-ups are 2nd-bast, four-seam fastballs are worst) and there appears to evidence of slight trends, it seems contact quality by pitch type is subject to a lot of noise. In other words, controlling for strikeouts and walks indeed levels the playing field among the pitch types.

Moreover, given the noise related to xwOBAcon, I’m starting to suspect it might be the kind of metric that behaves less reliably than others. In the same vein as our collective caution of trusting a batting average on balls in play (BABIP), home run-to-fly ball rate (HR/FB), or strand rate (LOB%) that’s egregiously high or low, we may need to exercise caution in trusting xwOBAcon marks that are egregiously high or low. We marry ourselves to Statcast’s “x” metrics as being the be-all, end-all validation of performance, yet xwOBA (and, perhaps more so, xwOBAcon) are moving targets in and of themselves.

Take Wheeler, for example. Here are his xwOBAcon marks by year, compared to the league:

Zack Wheeler xwOBAcon
Pitch Type 2018 2019
Pitch Type Wheeler League Wheeler League
Four-Seamer .357 .388 .387 .409
Slider .280 .343 .401 .355
Two-Seamer .242 .360 .393 .370
Splitter .403 .339 .324 .324
Curve .286 .345 .320 .354

I highlighted three specific pitches from Wheeler’s arsenal: his slider, two-seamer, and splitter. In 2018, Wheeler’s slider and two-seamer performed absurdly well (suspiciously so!) relative to the league-average xwOBAcon, while his splitter performed miserably. Fast-forward to 2019, and Wheeler’s slider and two-seamer have regressed beyond the league-average (if the pendulum swings to, it will also swing fro), while his splitter is now performing right at league-average.

Frankly, I think this says a lot. It’s one thing to look at a pitcher’s standalone xwOBA, compare it to his wOBA, and bet on over- or under-performance moving forward. It’s another thing to isolate the contact quality and identify exactly where and how he might be over- or under-performing, especially if contact management is not as much of a skill as we make it out to be. The table above could’ve been enough to determine the magnitude of Wheeler’s regression from his excellent 2018 campaign to this year’s underwhelming 2019 performance. It’s not that Wheeler’s pitches are unable to be better (or worse) than league-average — they certainly can be — just that they probably are unable to be better (or worse) than league-average by such wide margins.

There will be pitchers with offerings that consistently over- or under-perform the league-average xwOBAcon, just as there are pitchers who consistently betray the league-average BABIP, HR/FB, LOB%, etc. Ultimately, each pitcher should be evaluated individually, their trends over time parsed and assessed systematically. But I have a suspicion that year-to-year reliability of xwOBAcon, at only the pitch level but also (especially) at the pitch type level, is fleeting.

To be investigated another day. In the meantime: Throw a splitter, folks.

Currently investigating the relationship between pitcher effectiveness and beard density. Two-time FSWA award winner, including 2018 Baseball Writer of the Year, and 5-time award finalist. Featured in Lindy's Sports' Fantasy Baseball magazine (2018, 2019). Tout Wars competitor. Biased toward a nicely rolled baseball pant.

newest oldest most voted
Mario Mendoza
Mario Mendoza

Great work!