Early-Season Pitch-Modeling Standouts

This offseason, FanGraphs got some new stuff. More precisely, we got some new ways of measuring stuff, and command, and pitching overall, via pitch-level modeling. You can read about PitchingBot here and Stuff+ here. They’re really cool! Pitch modeling is a wonderful tool to both verify the eye test – that nasty-looking slider you saw, it’s actually nasty – and to find new pitchers to keep an eye on. Sure, strikeout rate and ERA and FIP can do that too, but stuff is a purer signal, because it’s entirely in a pitcher’s control. There’s no question of whether a hitter spoiled a great pitch, or whether that ball should have been a home run. There’s only the pitch, with its movement and velocity and release point.

Eno Sarris, the proprietor of Stuff+, has written about how quickly that model stabilizes, but for our purposes, let’s just say this: these pitch modeling tools give a great early look at which pitchers are working with the best tools early in the year. That doesn’t mean that they’ll all be great – they might not wield the tools in the correct order, or they might struggle with command, or they might wear down as the season goes on – but it does mean that they’re starting with an advantage.

I’d caution you against using these with excessive granularity this early in the season. If a pitcher’s Stuff+ has declined from 119 to 116, or if your team’s swingman has vaulted two points above the fifth starter, there’s probably not much signal in that. Instead, I’m going to paint with a very broad brush. I’m going to look at three groups of two today: two pitchers who both models agree have great stuff, two pitchers who both models are down on, and two where the systems disagree.

Let’s start with the good stuff.

Shohei Ohtani, Los Angeles Angels

PitchingBot Stuff: 74, 1st among starters
Stuff+: 152, 1st among starters
Oh, sure, that seems fair. We all already knew that Ohtani was great, but c’mon, he has the best stuff of any starter in baseball now? It’s true, though: he just keeps adding weapons, and the sweeping slider he throws might be the best in baseball. Just ask teammate Mike Trout, who waved ineffectually at one to end the World Baseball Classic. Or ask the A’s, who saw 45 of them in his Opening Day start.

As always, Ohtani’s splitter is magnificent. It’s been his best secondary pitch for years, and both models have it on par with various splitter specialists at the very top of the charts. This year, Ohtani has done one better, though, by improving his already-excellent fastball. He’s throwing it harder and yet getting more vertical movement. Oh yeah, he commands it too. The result is a pitch that’s gone from solid to overwhelming. Pity Ramón Laureano, who drew Ohtani’s maximum-effort ire, painted low and away with runners in scoring position:

Ohtani already had dominant stuff. He was already one of the best pitchers in baseball. Now he might be better than he was last year – when he finished fourth in Cy Young voting and hit 34 homers with a 142 wRC+. What can you do other than stare at his statistics in awe, mouth agape?

Dustin May, Los Angeles Dodgers

PitchingBot Stuff: 57, 16th among starters
Stuff+: 118, 15th among starters
May looked like the next great Dodgers starter at the beginning of 2021, but he suffered a UCL tear only five starts into the season. He barely pitched last year between rehab and a new back injury. I won’t fault you for taking a wait-and-see approach with him. That said, he looked electric in his first start of the year and both pitching models agree with the eye test.

The biggest change in May’s game is that he’s throwing a four-seamer more frequently than a sinker these days. He threw 36 four-seamers and only 18 sinkers in his first start, a big change for a guy who came into the big leagues as a sinker-only pitcher. Both of the models love both of the fastballs, which is a very fun sentence to write. His four-seamer has the requisite ride to make it hard to get a bat on, while his sinker is still a bowling ball with tremendous horizontal movement. It’s worth keeping an eye on his release points – he consistently releases the four-seamer three inches or so higher up than the sinker, which hitters might figure that out eventually – but for now, an upper-90s mix of two good fastballs is a great building block.

But wait – there’s more. May’s breaking ball, which is either a curve or a slider depending on who you ask, looks downright beastly. He throws it 85 mph on average, but with the kind of movement that used to get people accused of witchcraft. Poor Christian Walker never stood a chance on this one:

As a complement to that fastball/breaking ball mix, May throws a cutter that works mainly to keep lefties honest. It’s not the star of the show – it’s very clearly his fourth-best pitch – but the two pitch-level models agree that it’s above average as well. That’s a ton of good pitches, if you’re keeping track at home. If May stays healthy and pitches like this all season, the Dodgers will have an even better rotation than I expected.

Now let’s look at two pitchers who fare less well by the models.

Josiah Gray, Washington Nationals

PitchingBot Stuff: 31, 149th (out of 149) among starters
Stuff+: 71, 146th among starters
I’m not gonna mince words: Gray’s first start of the season was horrendous. If you think a 9.00 ERA is bad, you’re going to be horrified by a 10.63 FIP and three home runs allowed. It’s the kind of start that plagued Gray all too often last year, when he allowed 38 homers to “lead” baseball. One way to give up a lot of homers? Don’t have good stuff. Our two models seem to think that’s the case here.

Last year, Gray led with his fastball and complemented it with sliders and curveballs. This season, he’s trying out a new sinker and cutter, both at the expense of four-seamers, which means his two most frequently thrown pitches are the curve and the slider. That strategy has worked for a lot of pitchers, but there’s just one problem here: Gray’s two breaking balls aren’t very good. Whoops.

To me, the slider is the bigger offender of the two. It’s arrow-straight, the kind of gyroscopic slider that works when thrown in the low 90s to complement a big four-seam fastball. Gray throws it in the mid-80s, though, and throws it with backspin, to where instead of diving off the table, it hangs up enticingly. That’s a worst-case scenario because when he misses location, the pitch is static horizontally and fails to drop below the zone. That’s a great way to surrender a home run. Like, say, on an 0-2 pitch:

Gray’s curveball isn’t quite as offensive, though it’s still far too straight for my liking. The good news is, he throws it only three ticks slower than his slider but gets an additional 13 inches of downward movement on it, which means it’s far less homer-prone. But his cutter, which was responsible for the other two home runs, needs an overhaul too. Cutters are at their best when they have a bit of glove-side break, or maybe no horizontal movement whatsoever. Gray’s bends ever so slightly arm side, like a sinker that doesn’t sink or tail, and that’s a great way to end up on someone else’s highlight reel.

You might think I mentioned all these bad breaking pitches because Gray’s fastballs are good, but that’s not even the case. His four-seamer was already marginal last year and it looks meaningfully worse this year; he’s lost three inches of vertical movement. He also can’t command it; he threw 14 of them, 10 of which were balls. Let’s call it like it is: Gray doesn’t look like a major league pitcher right now.

Noah Syndergaard, Los Angeles Dodgers

PitchingBot Stuff: 37, 137th among starters
Stuff+: 77, 141st among starters
Lest you think the Dodgers are all stuff success stories, Syndergaard’s pitch mix didn’t look so hot despite an excellent first outing against the Diamondbacks. His velocity is down yet again, to 92.9 mph on average, a far cry from the days when he cosplayed as Thor, the god of thunder. His sinker, which he used far more frequently than his four-seamer, is cookie-cutter, with decent horizontal movement, decent sink, and decent velocity. Both pitching models think it’s slightly below average as a result, because throwing a pitch with no standout characteristics whatsoever isn’t a great idea against professional hitters.

That standout characteristic used to be Syndergaard’s velocity and four-seam rise, but those days look to be over. That also takes some of the starch out of his breaking ball, a 90 mph cutter/slider thing that looks excellent off of an unhittable fastball but less imposing these days. He’s getting a ton of ride on it, similar to the issue Gray has with his slider, though Syndergaard’s is at least thrown faster, giving hitters less time to salivate before they start swinging.

One big caveat here, though: Syndergaard located extremely well, which cures a lot of ills. I’m less well-versed in how the pitch-level models handle location, but Syndergaard gets a 67 command grade from PitchingBot and a 116 Location+ score, both of which are comfortably above average. It’s okay to throw subpar stuff if you’re putting it in the right places. Maybe that’s Syndergaard’s new plan. I’ll be watching with great interest, because it’s not every day you see someone take the career arc he’s followed, and I think it would be very cool if he succeeded after losing his Asgardian powers.

Finally, let’s indulge in a little pitch modeling debate.

Kodai Senga, New York Mets

PitchingBot Stuff: 58, 15th among starters
Stuff+: 98, 69th among starters
This comes down to what you think of Senga’s ghost fork, the splitter/forkball/changeup thing that garnered nine of his 10 swinging strikes in his opening start. It absolutely befuddled the Marlins, and Senga knew it: eight of those nine swinging strikes resulted in strikeouts because he waited for two-strike counts and then pulled the string on them. In my eyes, it’s his standout pitch.

Just don’t tell the pitching models that. PitchingBot simply doesn’t give the pitch a score, because Statcast lists it as a forkball and there aren’t exactly many comparisons for that. Stuff+ gives it a 76, where 100 is average, and uh, what? It must be a difficult pitch to model, but I don’t really love either system’s output here. I’m going to need to see more ghost forks and more starts to develop a better idea of what I think about it. For now, I’m considering it his best out pitch by a mile.

One thing the models agree on: the rest of Senga’s pitches are solid. His fastball plays well, combining velocity with solid shape. I’m not sure it would work that well if it weren’t consistently hitting 96-97 mph, but at that speed, it looks quite good to me and to the models. He throws an excellent sweeper, too, with a ton of horizontal break and a bit of drop. It’s slow enough compared to the rest of his pitches that hitters have a bit of time to adjust, but it moves enough that their adjustments often aren’t enough.

Senga also throws a cutter that PitchingBot thinks is slightly above average and Stuff+ thinks is slightly below average, but the key thing to look for here is the ghost fork. What you think about that pitch has a lot to say about what you think about Senga’s overall stuff. I don’t trust either model here, at least yet; I’d say go with your gut.

Aaron Civale, Cleveland Guardians

PitchingBot Stuff: 39, 128th among starters
Stuff+: 116, 18th among starters
Now here is a disagreement. PitchingBot thinks Civale has almost no stuff to speak of. Stuff+ thinks that he’s one of the nastiest starters in the game. What in the world is going on here? In a word, it’s everything. The two systems disagree across the board on how effective Civale’s raw pitch metrics are.

Let’s start with his two fastballs. They’re not going to grade out incredibly well by either measure, sitting in the low 90s without extreme movement, but PitchingBot hates them. It gives them grades of 26 and 28 on the 20-80 scouting scale. Stuff+ is more sanguine; it checks in at 76 and 101 on a plus-stats scale. That’s not a huge deal, because Civale mostly relies on a cutter, but it explains some of the disagreement.

Both models like Civale’s cutter just fine, though PitchingBot thinks it’s slightly below average while Stuff+ thinks it’s slightly above average. The biggest disagreement of the whole bunch is in his curveball, which everyone agrees is good. The question is how good. PitchingBot thinks it’s a 60, comfortably above average, but Stuff+ happens to think it’s the second-best curveball in baseball, behind only Taijuan Walker’s. I can see it; it’s a truly fearsome hook. It gets both more drop and more horizontal movement than average, and it’s not even that slow for a pitch with that much movement. Here, watch Tommy La Stella chase ghosts:

On the other hand, that’s the only swinging strike that Civale garnered out of 20 curves, and he also only landed one for a called strike. You might be tempted to write the pitch off given that. But this is one great strength of pitching models: they look at the raw building blocks rather than what hitters did against the pitch on a single day. Civale’s curveball looks mostly the same as it did last year, when it had a 19.7% swinging strike rate. It looks mostly the same as it did in 2021, when it had a 14.8% swinging strike rate, and in 2020, when it had a 19.7% swinging strike rate. The Mariners did a great job against it, but that doesn’t mean it’s not a good pitch. It’s reasonable to expect good season-long numbers given that we have a mountain of data suggesting that this pitch shape works for him.

The truth on Civale is probably somewhere between the two models. I’m sympathetic to them here; pitchers who throw cutters as their primary pitch present problems given that differential from primary pitch is an important input. I’ll direct your attention to Graham Ashcraft, who is fourth in Stuff+ and 27th according to PitchingBot’s Stuff grades, as a similar case. Modeling baseball pitches isn’t easy. That doesn’t mean that it’s not useful, though, and I hope this was a helpful look at the way our leaderboards work.