When the Cardinals played the Marlins on Wednesday night, two of the hardest throwers in baseball made appearances. Jordan Hicks was flawless, getting two strikeouts and four grounders in two perfect innings. Tayron Guerrero pitched a 1-2-3 inning, but with a tad more excitement than Hicks produced. He allowed a fly ball to medium-deep center field, and fly balls are always adventures given the current state of home runs. As I listened to the game, however, the announcers were quick to mention that Guerrero wasn’t in great peril with that fly ball, because it’s hard to hit home runs off of someone who throws so hard.
My statistical curiosity was piqued by that comment. It’s something I’ve heard from time to time, and it seems logical — I’ve watched a fair amount of Hicks appearances in the past two years, and batters seem tremendously uncomfortable when facing him. On the other hand, there are plenty of things I’ve heard about baseball that seem logical but aren’t true. I grew up knowing when the right time to bunt was and how some batters were just better at hitting in the clutch, and those have since been proven false. What’s to say that “throwing harder suppresses home runs” isn’t just another in a list of untruths?
To a certain extent, every time you use xFIP to describe a pitcher’s skill level, you’re ignoring this pearl of broadcaster wisdom. After all, if you’re regressing everyone’s home runs back to a league-wide average, that implies that no one has special skills to suppress home runs when the ball is hit in the air. No one would say that xFIP is a perfect and foolproof predictor, but it does do fairly well when it comes to ERA estimators — it beats FIP and ERA, for example.
As a general believer that skill is over-ascribed in baseball (not every 2.5 ERA or 150 wRC+ is hiding a great process — sometimes it’s just a hot stretch), I’m naturally inclined to go with xFIP’s explanation of how fly balls become home runs. That’s not to say that throwing hard doesn’t have advantages, obviously — baseball’s ever-creeping velocity increase is proof of that. It helps with swinging strike rate, of course, and therefore directly helps increase strikeouts, the most valuable thing a pitcher can do. When the ball is struck, however, the hitter has, necessarily, not swung and missed. In fact, you’ve probably heard people say that when a hard thrower allows contact, it’s usually harder than normal because the ball was coming in so fast. Wouldn’t this increase home runs per fly ball? Sounds like something worth looking at.
I looked into the fringe benefits of throwing hard last year, but I wanted to do a more focused study of HR/FB rate. Looking at whether velocity and ERA are correlated, for example, doesn’t really tell us much. There are all kinds of selection biases in play here — if you throw 88 mph on average and are a major league pitcher, you’re probably doing something else amazing. Let’s ignore broad pitching metrics then and specifically focus on one thing: does throwing harder lead to home run suppression?
To answer this question, I went with the most straightforward experimental design I could think of. I looked at every pitcher who has thrown at least 50 innings between 2015 and the present. This gave me a 757-pitcher population, which I trimmed to 755 by eliminating knuckleballers. Then I looked at their average fastball velocity and HR/FB rate. There was one key question here — should I look at overall HR/FB, or focus on what happened to fastballs only? I settled on looking at overall rate for one main reason. I’m looking for useful takeaways here, not curiosities. If higher velocity decreases fastball home run rate but increases breaking-pitch home run rate such that the overall rate is unchanged, that’s not really useful. I want to know whether I can look at a single radar gun reading from a pitcher and say something meaningful about his home run suppression skills, and that means looking at overall rates rather than pitch-specific ones.
With that in mind, I did the simplest test I could think of — I regressed HR/FB against average fastball velocity. The results were not promising. Here, take a look at some dot soup:
This is a whole lot of nothing. Maybe — maybe — you can perceive a general downward trend in the data. It’s swamped by noise, though. The r-squared of this relationship is a don’t-even-bother-reporting-this .015. If you used velocity to predict HR/FB rate, you’d be capturing mostly noise, and you wouldn’t even get very interesting predictions. Use that fit, and Hicks has the lowest HR/FB projection in baseball at 11.6%. Jered Weaver’s 14.3% is the highest. Predicting league average for everyone does almost exactly as well when it comes to the standard error of predictions.
Well, that’s it then, right? I came up with a specific test I wanted to run, and the data shot me down. Study and grow stronger, and so on. Not so fast, my friend. Looking at whether velocity is generally correlated to home run suppression isn’t exactly the question I was trying to ask. The conventional wisdom we’re looking at here is that throwing abnormally hard provides HR/FB benefits. The scatterplot of velocities and home run rates suggests that going from 91 mph to 92 mph doesn’t do much of anything for you, but we don’t care about the vast majority of the data. We only want to look at the real freaks, after all — the Hickses and Guerreros of the world.
Let’s look at our data a separate way. Instead of one giant population of pitchers, let’s split it into two. On one side, we have a tiny group, the 25 pitchers with the fastest fastballs in our sample. They’re not all Hicks and Aroldis Chapman types — Carlos Estevez makes an appearance, as does the aforementioned Guerrero. They’re hard throwers, but not universally great. Our second group is everyone else, the 730 pitchers who aren’t among the very fastest throwers in baseball.
Our first group sports a better-than-average 10.4% HR/FB rate. That doesn’t exactly scream outlier, but the overall average HR/FB rate since 2015 is 12.9%, and the average of our sample (pitchers with 50 or more total innings pitched) is 12.8%. We’re not talking a small sample, either — our group of 25 velocity standouts has more than 5,000 innings pitched over the past five years, as many innings as Nolan Ryan pitched in his entire career. Maybe we’re onto something here.
To evaluate whether there’s anything real, I performed a one-sided Student’s t-test, which attempts to measure the likelihood that two different samples come from the same underlying population. In English, this test looks at what the odds are that there’s just one mean HR/FB rate for all pitchers, and that the two groups look different by random chance. Well, we’re in luck: the t-test shows us a result. Are you a fan of long decimals? The p-value of this test is 0.000072. In other words, there’s a 0.0072% chance that the high-velocity group would have this low of an HR/FB rate or lower if the two groups have the same underlying mean. This isn’t to say that they’re definitely pulled from different populations, but it’s a pretty strong result. The hardest throwers really do seem different.
Now that I’ve come to a conclusion, allow me to walk it back somewhat. This, to me, is strong evidence that true outliers in terms of fastball velocity help suppress home run rates on fly balls. If you merely split the population down the middle and test whether those groups have a different mean, the t-test is inconclusive. However, the hardest throwers likely really do have lower true-talent HR/FB rates than the population of mere mortals. That said, we’re not exactly talking about a huge effect here. The fastest throwers had a 2.4% lower HR/FB rate than the sample as a whole. That’s just not a whole lot of home runs.
Take Guerrero as an example. He’s the pitcher who started this whole train of thought, after all. Over his entire 88.1-inning career, he’s allowed 82 fly balls. Apply that 2.4% difference in home run rate, and that gets us to two home runs less he “should” have allowed by throwing so dang hard. If we modify the xFIP formula to use the HR/FB rate for the high-velocity cohort rather than the league rate as a whole, that moves his xFIP lower by roughly .3, from 4.55 to 4.25. It’s an effect, in other words, but not a huge effect. For an extreme sinker pitcher like Hicks, who has allowed only 41 fly balls in his career, the effect is even smaller — he gets a .125 xFIP adjustment.
What does this all mean? Well, the announcer mostly had it right. It really is harder to hit home runs against pitchers who throw really, really ridiculously hard. On the other hand, the announcer mostly had it wrong. Have you ever tried to think about the difference between a 10.4% chance of something happening and a 12.8% chance? Watch literally a hundred observations, and two will be different. That’s simply too fine of a distinction for our brains to make sense of. If someone throws absolute gas, fine, they’ll probably allow less home runs. Don’t take that relationship too far in your head though. The vast majority of home run suppression is still random chance. The next time you find yourself thinking that a pitcher deserves a 5% or 6% HR/FB, let this temper your conviction. Differences exist, but their magnitude isn’t what you’d hope. In fact it’s almost too small to perceive.
Ben is a contributor to Fangraphs. A lifelong Cardinals fan, he got his start writing for Viva El Birdos. He can be found on Twitter @_Ben_Clemens.