When Aaron Nola’s curveball is on, it’s one of the best pitches in Major League Baseball. Among pitchers who’ve thrown at least 5,000 pitches since the start of 2017 (106 in all), Nola’s curve generated 543 swings and misses. That means nearly 6 percent (5.9%) of Nola’s entire pitch total over the last three years is made up of whiffs on curves, a huge total that has him tied with Charlie Morton for third-best percentage in the league in that time. Only Arizona’s Zack Godley (7.5%) and the recently traded Corey Kluber (6.1%) rate higher.
But 2019 didn’t feature Nola’s peak curveball as often as we might have hoped. Year-over-year, its performance took hits (literally and figuratively) in a few areas:
- Whiff% dropped from 41.1% to 34.4%
- K% dropped from 41.5% to 33%
- wOBA allowed jumped from .193 to .262
- SLG allowed jumped from .249 to .328
And all of that happened as its usage rose from 31 percent to 35 percent, magnifying the increased damage output. It is, as the new skipper would say, not what you want.
But...why, though? Why did such a lethal pitch suddenly become more hittable in 2019?
Warning: Numbers and graphs and things, oh my!
The reason, it seems, largely boils down to movement.
Nola’s curveball often had more lateral, sweeping movement in ‘18 than it did in ‘19, where it straightened out and took more of a 12-6 form than it typically had before.
Take, for instance, two particular starts, selected somewhat arbitrarily. The first, shown on the left in the graphic below, was one of Nola’s best starts of that bronze medal Cy Young season: 7 IP, 3 H, 1 R, 1 BB, 11 K, and 21 swinging strikes against the Mets. The second, on the right, was Nola’s final start of the 2019 season: 5.2 IP, 5 H, 4 BB, 9 K, and 1 HR allowed, but with 17 swinging strikes (tied for his most in a start last year).
Nothing really links these starts other than being later in the year with good K results and high swinging strike totals. They’re put side by side here to paint a picture and tell a tale of two different pitches.
For one, notice the location. Nola spotted far more curves over the middle and arm-side half of the plate than he did in his dominant ‘18 start. More importantly, though, look at the shape of the pitch tracking tails, the arcs of the lines tracing from release point to the front of the plate. The perspective doesn’t change from picture to picture, and yet it sure seems like the curves on the left are moving more to the glove side than the ones on the right. It’s not an illusion!
The curve lost, on average, an entire inch of horizontal movement. No other pitch saw nearly as drastic a change in either direction.
But all of that might only matter if the reduced movement led to worse outcomes for Nola. It’s tougher to parse outcomes by pitch with movement on the BrooksBaseball site (as far as I can tell), but Statcast data allows you to piece these things out, pitch-by-pitch. The difficulty we come across in braiding those two datasets is that they diverge and measure pitch movement differently; Statcast measures of horizontal movement (pfx_x in their data exports, if you find yourself looking for a fun time with spreadsheets) average about 4 inches of extra movement in their measures. That doesn’t line up with the Brooks plot above, so we’ll keep it relative: As movement increases or decreases, do we see a noticeable change in outcomes?
Given what we know about Nola’s curveball from when it was at its regular best (2018), we know its flight path takes it away from right-handed batters. It dives away toward or beyond the outer edge of the zone, looking like a tantalizing pitch and often inducing swings that eventually stand no chance of making contact.
I mean, c’mon. What’s Xander Bogaerts supposed to do with that, other than what he did?
Compare that curve to Bogaerts in 2018 with this one to J.D. Martinez last August. Both in Boston, both with the same camera angle from center field. See if you notice a difference.
Hopefully, the difference is perceptible, but rest assured: These are different curveballs. Statcast data says the curve to Bogaerts in 2018 had about 11 percent more horizontal movement and about 22 percent less vertical movement than the curve to Martinez in 2019.
This is far from airtight data science, so those of you who might be professionals are excused for that bit of cringing you’re doing, but taking a closer look at the registered movement of these curves and the outcomes that came with them should give us an idea of whether this hypothesis holds up, or let us know if we’re on the right track.
That is...hardly the conclusive evidence we were hoping for. The histograms above show that, purely based on horizontal movement, Nola did not see a significant difference in whiff versus in play outcomes. In a way, that’s to be expected, right? More things go into a hitter’s ability to make contact than purely horizontal movement: Whether they identify it in flight and at what point that happens, the pitch’s ultimate proximity to the strike zone, etc.
The histograms above also lump all balls in play (outs and hits alike) into a single bucket. Nola could still get a favorable outcome with a ball in play: A tapper, a pop-up, a soft liner, any of which would be a “good” outcome for a pitcher, even if they turn into a lucky hit. Somewhat. We’ll consider them “good” outcomes for this next exercise, since I’m not content debunking this thought based on the graphs above.
As Nola’s movement increases, does the likelihood of a “good” outcome increase? That is, do we see a higher probability of swinging strikes or weaker contact as the pitch darts around more?
Over the past three seasons, Statcast has tracked 2,986 Aaron Nola curveballs. From that sample, we have a few baseline readings.
- Median Horizontal Movement: 1.3571 feet
- 90th Percentile Horizontal Movement: 1.7298 feet
- 75th Percentile Horizontal Movement: 1.5212 feet
- 10th Percentile Horizontal Movement: 1.0936 feet
Using those values as benchmarks, we can construct a quick table to see if good outcomes increase as horizontal movement increases. “Good” outcomes are defined here as swinging strikes or a ball in play with an exit velocity of 85 MPH or slower. Bunts are excluded, because that would be cheating.
Nola vs. RHB, Outcomes by Horiz. Movement (2017-19)
|Movement||Count (of 2,986)||Good Outcome %||In Play 85-||Whiffs|
|Movement||Count (of 2,986)||Good Outcome %||In Play 85-||Whiffs|
|1.0936 ft. or more||2,687||31.0%||351||483|
|1.3571 ft. or more||1,493||31.0%||190||273|
|1.5212 ft. or more||747||32.0%||89||150|
|1.7298 ft. or more||299||36.1%||38||70|
Well, that’s not nothing. There’s an upward trend, with a huge surge in good outcomes for pitches with horizontal movement in Nola’s 90th percentile. More swerve sure seems better, based on this.
The 2017 season saw the majority of these side-to-side-type curveballs get thrown, with a steady decrease over the next two seasons: Nola threw 146 curveballs with more than his three-year sample’s 75th percentile mark in 2018 compared to just 55 in 2019.
Opponents’ In-Zone SLG vs. Nola’s Curveball (2017-19)
Something changed. Some type of adjustment was made that altered the shape of Nola’s curveball, and it worked to his detriment in 2019. Perhaps it’s related to the change in baseball composition, but that was present in 2018, too. It could be a grip change, or the change in his release point that has him about an inch closer to his body.
The cause is unclear. I know I can’t say for sure what’s contributed the most to the change in the curveball’s shape. What I do see is that the way Nola used to throw the pitch was extremely good, and the 2019 version of it was less so. In the grand scheme of things, even the 2019 version of his curveball is a good pitch. But rediscovering the sweet spot he’d found in 2017-18 should go a long way toward reestablishing Nola’s curve as the truly elite pitch it should always be.