Squared-Up Rate and Launch Angle: A Visual Investigation

I continue to find Statcast’s bat tracking data fascinating. I also continue to find it overwhelming. Hitting is so complex that I can’t quite imagine boiling it down to just a few numbers. Even when I look at some of the more complex presentations of bat tracking, like squared-up rate, I sometimes can’t quite understand what it means.

I’ll give you an example: when I looked into Manny Machado’s early-season struggles last week, I found that he was squaring the ball up more frequently when he hit grounders than when he put the ball in the air. That sounds bad to me – hard grounders don’t really pay the bills. But I didn’t have much to compare it to, aside from league averages for those rates. And I didn’t have context for what shapes of squared-up rate work for various different successful batters.

What’s an analyst to do? If you’re like me in 2024, there’s one preferred option: ask my friendly neighborhood large language model to help me create a visual. I had an idea of what I wanted to do. Essentially, I wanted to create a chart that showed how a given hitter’s squared-up rate varied by launch angle. There’s a difference between squaring the ball up like Luis Arraez – line drives into the gap all day – and doing it like Machado. I hoped that a visual representation would make that a little clearer.

First things first: I downloaded every ball in play from this year where Statcast recorded a bat speed, pitch speed, launch angle, and exit velocity. Then I manually calculated whether each batted ball was squared up. As a refresher, a batted ball is squared up if the ball travels at 80% of its maximum theoretical velocity, as measured by a proxy formula: 1.23 * bat speed + 0.23 * pitch speed at home plate, which is roughly 92% of pitch speed at release. If you’re interested in following along with me at home, you can find that data here. If not, bear with me, because I’d like to show you some pictures I made.

From there, Gemini (my LLM of choice, though I’m sure others would end up in roughly the same place) and I got to work. We calculated the squared-up rate of each hitter at each angle. I had to make a few decisions here about how to aggregate data. I decided to bulk up every angle by looking for balls hit within 10 degrees of it either way, then threw out every bucket that didn’t have at least 20 data points after doing that bulking up. There’s some overlapping data this way, but sample sizes are small enough, and I think that hitter intent is broad enough, that if you’re wondering how frequently someone squares up a batted ball at 15 degrees, looking at 5 degrees and 25 degrees are both useful inputs.

Those are the parts that I came up with, but I wasn’t quite sure how to turn that concept into a program that could make graphs out of my idea. But that’s nothing I couldn’t solve after a few hours of coming up with ideas, translating them into Python code using generative AI, finding problems with the code, coming up with new ideas to solve those problems, translating those new ideas into new code, finding new problems… you get the idea.

Essentially, I wanted a graph of how good Machado is at squaring up the ball depending on whether he’s hitting it down, flat, or up. Great news. I got exactly that graph:

He’s squaring up a ton of his contact on the ground, just like we knew. He’s getting the most out of his bat speed far less frequently at the juicy launch angles in the 20 degree range. That doesn’t sound much like Arraez, the bat control god, at all. But what does Arraez’s graph look like? It looks like what you’d expect:

As a side note, the size of the circles is proportional to the percentage of contact in that bucket. Arraez’s biggest circles are line drives of various types. He hardly has any extreme grounders or extreme popups. That’s what amazing bat control looks like.

How does that compare to Machado? After a round or two of dancing with Gemini, the tool I built can help with that too:

You could pretty much guess this even before this graph, but it’s still nice to see it in pictures. Machado is squaring up grounders at the same rate, but his swing just isn’t getting it done in the air right now. We can throw in a third hitter to show what it looks like when you’re the opposite of Machado. Here’s Bryce Harper, whose uppercut swing is etched into pitchers’ nightmares everywhere:

Harper is the new series, in Philly red. When he hits the ball on the ground, he’s rarely squaring it up. In other words, those are largely mishits; when he’s squaring the ball up, it’s generally in the air. He consistently beats Machado at squared-up rate in the air, and he hits more fly balls as well. He might not square the ball up as frequently as Arraez, but he swings much harder and connects often enough. Perhaps unsurprisingly, he’s mashing so far this year.

For another fun comparison, let’s look at Aaron Judge and Juan Soto:

They’re both making pristine contact across the board. They’re at or above an 80% squared-up rate for pretty much everything in the air, and they’re both swinging hard too. That’s a deadly combination. Judge is even avoiding grounders; he doesn’t really have a left tail to speak of. His biggest cluster of launch angles is the most dangerous one in baseball when you’re hitting the ball hard. In other words, he’s swinging hard, squaring the ball up frequently, and doing it on home run trajectories. No wonder he’s slugging .703.

Those two elite hitters are getting it done in almost the same way. But it’s not the same for everyone. The Dodgers’ three stars show some variation:

Mookie Betts has turned into an extreme fly ball hitter. Here’s the graphical evidence of why that’s working: He’s contacting the ball most squarely at around 30 degrees of lift, and he’s hitting the ball in the air incredibly frequently. To the extent that he has mishits, he’s getting too far under the ball and popping it up, which makes sense given his overall approach. Shohei Ohtani, too, is squaring the ball up most frequently in the air. He isn’t hitting a ton of grounders, though more than Betts. He’s also absolutely rifling low line drives — look at all those high blue circles in the 10-20 degree band.

Then there’s Freddie Freeman. He hits everything square at about the same rate. His most frequent launch angles are basically everything from 10-40 degrees. There’s almost no variation in his line; both Betts and Ohtani have higher highs and lower lows. Freeman’s swing seems to be a chameleon; it just changes to fit the contact type. In a lot of ways, he’s a burlier but less precise Luis Arraez:

They both just rake, plain and simple. Arraez hits it flush more frequently, of course, but Freeman swings 7 mph faster. Arraez focuses more on the 5-15 degree band; Freeman taps into his power by hitting more balls in the 25-35 degree range. But they’re both absolutely peppering everything, whether in the air or on the ground, and they both hit a ton of line drives. These guys are incredible.

We can do more. Want to see some young American League shortstop dynamos? Take a look at Bobby Witt Jr. and Gunnar Henderson:

Witt has a promising contact shape, but not a perfect one. It’s like Freeman’s, only shifted down a bit and with more grounders. There are some red flags, like his relatively low squared-up rate when he’s putting the ball in the air. To be honest with you, though, I’m not sure how important squared-up rate is in these small and cut-up samples. I’m more interested in shape for now, and I’ll have time to do more testing of how much the levels matter later. The key part, for me, is that Witt’s most frequent outcomes are fly balls and line drives, but his most frequent square-ups occur on grounders. Make that correction, and even more upside could be available.

Henderson, on the other hand, seems like he was designed in a lab. He squares the ball up most frequently at the launch angles where hard contact is most advantageous. He doesn’t have enough popups to get any dots up there. His grounders are all mishits. Sure, maybe he could concentrate even more batted balls around his best swings, but he’s doing exactly what I want every hitter to do: hitting the ball flush when he elevates, and doing so with plus bat speed.

Here’s a mystery that this data can solve: Why does Henderson have 20 homers to Witt’s 11? Witt hits the ball harder, hits fewer grounders, and even has a higher barrel rate. But Henderson’s swing is designed to square the ball up in the air more frequently, so he’s lined up high-value launch angles and high-value exit velocities better than his Kansas City counterpart.

I think this data will get far more interesting when we have access to multiple years of history. I’d love to know if Kyle Tucker’s swing shape has changed in conjunction with his lower groundball rate and otherworldly production. I’d be interested in seeing how hitters who change their batted ball tendencies change their squared-up tendencies. I want to see whether Nick Castellanos has always squared up the ball exactly like Bryce Harper, or whether he used to have a different shape and the new one is correlated with his downfall:

I haven’t quite figured out what to do with all of this in the long run. I think it’s more of a storytelling tool than something that can tell you who will be great and who will struggle. That said, I love the stories! Arraez is good in the ways you’d expect. Betts maxes out on power with his swing. Harper’s uppercut is cool to see in data. And how about that uppercut against Yandy Díaz’s ground-friendly ways:

Yandy is smashing those grounders. You and I already knew that, but it’s cool for this data to verify the eye test. That’s basically what this is to me; a way of converting some dry data points into a story.

The tool I built isn’t live on the pages of FanGraphs for many reasons. It’s hilariously rudimentary. It’s buggy. It’s programmed by me, a coding imbecile, rather than by our team of excellent developers. It might not even be useful in the long run.

So no, you can’t just click on a single link and play around with this to your heart’s content. But I have two things to offer that will hopefully make it up to you. First, this is an open source project. You can find the Python script that generates these graphs here, along with the underlying data. I’m certainly not confident that this is the most efficient way to do things – I was building from scratch without a lot of experience in this area. If you have some improvements or whatnot, let me know!

Second, I happen to have the code and the ability to post pictures to the internet. So if you’re interested in a particular comparison, ask me below in the comments. I’ll get to as many as I can for the next day or so, because I understand that “hey, just learn how to use this computer programming language real fast” isn’t exactly a way to guarantee broad access.

So, yeah. That’s the end! No real conclusion today, aside from a) I think this tool is cool and b) here are some pictures of it. I hope you like it, and I hope there will be more bells and whistles before long.