Throwing Hard is Good... Right?

I don’t have anything of incredibly great import to talk about today. Between John Gant’s pursuit of history and my farewell to Tommy Pham, I’ve gotten a lot of whimsy and sentimentality out of my system, so let’s get back to manipulating numbers and making broad claims that are only mostly supported by data! In all seriousness, there’s something I’ve been meaning to talk about here for a while, and now seems as good a time to do it as any. I don’t know how much this is true and how much I’m just projecting on other Cardinals fans, but to my mind there are two extreme takes about Jordan Hicks this year, both of them wrong. The first is what I’ll call the velocity truthers. This is a pretty easy one- Jordan Hicks throws the ball so dang hard. How could he not be great? He’s got a shiny ERA too, and it used to be even shinier. Great pitcher. Let’s move on. I don’t think this view captures the full Jordan Hicks experience.

Neither, though, does the opposing view, which I’ll call maximum belief in broad statistics, or the xFIP truthers. This view is slightly more dressed up, but still pretty reductive. It goes basically like this: Hicks doesn’t strike out that many, or walk that few, so he can’t be a great pitcher. This school will often mention xFIP, a stat that looks at only strikeouts, walks, and fly balls to figure out how well a pitcher is doing. To listen to this side’s argument, Hicks has benefited from few of the fly balls he’s allowed leaving the park. If you regress that number to league average, he’d be a pretty generic reliever. For reference, league average xFIP is 4.12, and Hicks is running a 4.18 xFIP. What’s all this fuss about this below-league-average pitcher, this school of thought goes. The Cardinals should take their found money and move on with life, not keep running him out there in high leverage spots.

Maybe I gave it away with my introduction, but I don’t think either of these views is right. To be sure, I’m willing to agree with parts of both of them. For example: Jordan Hicks does throw hard. That’s just true. On the other side, he really doesn’t strike out that many guys. That’s just true. He’s struck out just 19.3% of the batters he’s faced this year, and even if you exclude his whiff-light April when he was adjusting to the major leagues, that rate only jumps to 23.5%. Simply put, that’s not a ton of strikeouts. I can accept that both of these things are true. There are holes in these theories, though.

Let’s get the velocity-truther holes out of the way first. Look, yeah, he throws the ball hard. That’s not up for debate. That’s just not enough to succeed in the majors. It’s not a radar gun contest. If you want a demonstration of this, take a look at last year’s Arizona Fall League stats. Hicks was there getting in a little extra work transitioning to relief. In nine games, eight of them out of the bullpen, he compiled a 6.32 ERA. That undersold it, too- he gave up two unearned runs. Basically, he just wasn’t good, and throwing harder didn’t change that. You want some further evidence? The two hardest throwers in the league this year are Hicks and Aroldis Chapman. So far, so good. Here are some other names from the top ten: Tayron Guerrero, Joe Kelly, Tanner Rainey. Throwing hard isn’t some panacea for otherwise bad relievers. It’s nice, but you can bring heat and still be bad. Pure velocity isn’t enough. ERA isn’t really enough either. Hicks is running a 3.1 ERA on the year, but c’mon, it’s 58 innings. I need to see more than that to be totally bought in. Plenty of people put up random stretches of good relief.

The argument that Hicks is below-average by regressed metrics is the one I really want to talk about today. It’s a pretty easy argument to make. “Hey, look at this metric. It has an X in the name. That stands for expected! Hicks is worse than average in it. He’s not good.” That’s all you need. Here’s the thing though- that’s too reductive. If the velocity truthers ascribe too much uniqueness to Hicks, the xFIP truthers ascribe too little. Over time, if you know nothing about a pitcher, it’s probably safe to assume that he’s going to allow home runs at about a league average rate. In the end, if you assume that about everyone, you’ll be right on average! Definitions are fun that way. The thing is, though, we don’t know nothing about Jordan Hicks. We know one very relevant thing- he throws the ball harder than anyone else in baseball. I’m no baseball scientist, but I’d naively assume that matters. To figure out whether there was anything behind my assumption, I took a sample of every reliever with at least 50 innings pitched from 2014 to the present. I took their average fastball velocity, ERA, FIP, and xFIP. Then, I looked for relationships between the variables. First off, let’s look at velocity and ERA:

Yeah, not so much. Maybe there’s a little something there, but it’s hard to tell. This makes some intuitive sense- as discussed earlier, throwing hard isn’t a cure for all ills. Joe Kelly exists. Still, though, we’re not interested in ERA. We’re interested in whether Hicks is better at beating his peripheral stats than the average pitcher. To test that, I decided to look at the relationship between fastball velocity and the gap between a pitcher’s ERA and his xFIP. That sounds like a mouthful, but here’s the thinking: xFIP makes everything about a pitcher other than his strikeout, walk, and groundball rates be league average. If pitchers who throw really hard really do give up less home runs, xFIP should systematically underestimate them. As a group, they should have comfortably lower ERA’s than xFIP would imply. So, how does that look?

Honestly, I can’t really see much more in this. There’s a lot of noise in the middle, and not enough data points on either end. Luckily, though, I have a trick up my sleeve for data like this. I’m not actually concerned with whether throwing harder in general is good. I only care about whether truly elite velocity matters. To get a gauge of this, I simply looked at the top five percent of hard throwers. Let’s take a look at their stats, and also the stats of the whole sample:

This is a bit more compelling. As a group, pitchers who are at the very top of the charts in fastball velocity are better pretty much across the board. They run lower ERA’s and beat their peripherals by more than the group as a whole. Still, I wanted to go further. I performed one-sided t-tests on all of these statistics relative to the rest of the population. A quick recap for people who don’t have a lot of occasion to break out measures of statistical significance: a one-sided t-test gives you the likelihood that one set of data (in our case, the hard throwers) is lower (or higher) than a separate set of data (the rest of the pitchers). The lower the number, the more likely it is that the tested sample is truly different. Since no one should teach statistics without giving at least one terrible rule of thumb, let me add that for some arcane reason, results lower than .05 are generally considered significant. How do our t-test values look?

Well, that’s a lot of significant things. Starting from the left, I’ll go through them quickly. ERA, FIP, and xFIP are all quite likely to be significantly different. I didn’t go out to enough decimal places, but trust me when I say it’s pretty unlikely that they have the same population mean as the rest of pitchers. Velocity is also different. That makes sense- given that we bucketed these pitchers by how fast they throw, it would be weird if that weren’t the case. The degree to which pitchers outperform their FIP (for completeness’s sake, an ERA predictor based on K, BB, and HR rates) is probably not meaningful. Put another way, if you know a pitcher’s three true outcome rates, you won’t gain any further insight into their ERA by adding how hard they throw.

Next, we arrive at the main attraction, ERA - xFIP. If the ERA - xFIP from the sample of hard throwers has the same mean as the population as a whole, there’s only a 2% chance we’d observe values this low by chance. To put it another way, if I give you a pitcher’s xFIP, you might make one guess about their ERA. If I told you they throw incredibly hard, though, you’d be right to revise your guess lower. In short, throwing really hard does seem to suppress home run rates. That’s shown in the last box, HR/FB. This one is statistically significant at basically all levels- pitchers who throw hard have fewer of their fly balls turn into home runs. Why is this? Well, I can think of one pretty likely reason. Have you ever watched someone swing against Jordan Hicks? What about Aroldis Chapman? Guys are just up there trying to get the bat on the ball. With the pitch coming in so fast, hitters tend to take reduced swings at the ball. That’s all well and good when you’re trying to put the ball in play, but it’s hard to hit a home run that way.

You want some bad anecdotal evidence after an entire article that tries to show real evidence? You’re in luck! Hicks allowed one home run across 60 innings pitched in 2016. He allowed three over 105 innings in 2017. Aroldis Chapman allowed seven home runs in 2013, and then promptly allowed one, three, two, three, and one home runs in his subsequent seasons. Two people does not a sample make, but hey, the samples are up above. We’re just at the part where I provide witty banter and relevant examples now.

Just for funsies, I performed one last calculation. I took Hicks’ results this year and plugged them into the xFIP formula, but I replaced the league-average HR/FB term in the formula with the average for his cohort, the relievers who are among the 5% who throw hardest. This is about as scientific as imagining what a lion would look like with eagle wings, but I’m the one writing the article, and we’re doing it. If that’s his true talent for preventing home runs, Jordan Hicks would have an xFIP of 3.88. This provides a little context for how well he’s pitched since the start of May too. Plug those stats into my velocity-adjusted predictor, and it spits out a 3.22 xFIP, basically this year’s Craig Kimbrel.

I’m writing this Monday night, and it’s getting late, so I don’t have a great pithy conclusion for you. Basically what I’ve got for you is this. Stop treating Jordan Hicks as absolutely unique. That’s asinine. People have thrown baseballs before. At the same time, stop treating him like he’s Joe Average. He’s not, and applying some one-size-fits-all formula to every pitcher in baseball is a great way to miss people with exceptional skills. It’s all well and good to regress things to league average when you don’t know any compelling information about somebody. When you do, though, use it! It makes analysis a bit harder than just looking at one number, but baseball analysis would be miserable if one number was all you needed. Luckily, there’s more to the game than that.