An Arm Angle Update That Ends With a Mystery

Early last year, I wrote two articles exploring the handful of pitchers who decided that, depending on the handedness of the batter they were facing, they should change not just their pitch mix but something more fundamental about themselves as pitchers. Some drastically lowered their arm angle against same-handed batters, while some scooched from one side of the rubber to the other. I mostly wrote about these pitchers because they were fun to watch, but I also dived into the reasoning behind their decisions. It wasn’t hard to understand what they were thinking: All things being equal, throwing from a lower arm angle works better against same-handed batters, while a higher arm angle works better against opposite-handed hitters. I even had numbers to back it up. I ran correlation coefficients between the pitcher’s wOBA allowed and their release point, and I used average velocity as a sort of control variable.

The correlation coefficients in this table are quite small, but they indicate that when the pitcher has the platoon advantage, vertical release point matters a whole lot. In fact, in that sample, it has the same correlation to success as velocity, which is definitely a surprise. When the batter has the platoon advantage, vertical release point doesn’t have any bearing on their success, but horizontal release point does. That’s why some pitchers scooch all the way over to the opposite side of the rubber.

Now that you’re all caught up, it’s time to address the big flaw in those numbers. The problem with my data was that I wasn’t actually using the pitcher’s arm angle. I was using their release point – literally the spot in the air above the mound where the ball leaves their hand – as a stand-in.

Here’s what I wrote at the time:

To this point, I’ve been using vertical release point as a proxy for arm angle, but it’s not that simple. Variations in height, wingspan, and mechanics mean that the equivalence between the two is imperfect. You can reverse engineer arm angle from PITCHf/x or Statcast data, but that will only give you an estimate. I imagine Statcast is capable of tracking arm angle. If that data ever becomes accessible to the public, I think it might significantly change our understanding of the platoon advantage.

As you may know, Statcast recently made arm angle data available to the public. Are you ready to have your understanding of the platoon advantage significantly changed? Well that’s too bad, because the results aren’t particularly shocking. However, we can still glean a few interesting facts. First of all, let’s update our table by adding in arm angle. Because the data only goes back to 2020, the sample size is a bit smaller, which in turn makes the correlation coefficients smaller.

Despite the smaller sample size, it’s reassuring to see that the trends are holding. Let’s start with same-sided hitters. A higher release point, and one that’s closer to the center of the rubber, works better against them. The new column tells us that against same-sided batters, when the pitcher has the platoon advantage, the actual arm angle is much more important than release point, which is what we would have been expected. Cutting out those confounding variables makes it clear that we don’t need to use release point as a proxy anymore; arm angle is what matters.

Now let’s turn to opposite-handed batters. When the batter has the platoon advantage, I was really surprised to find that horizontal release point had a slightly stronger correlation to wOBA than arm angle did. Once again, these are extremely small numbers, so we shouldn’t read too much into any of this, but if you wanted to, you could take this as evidence that the pitchers who scooch all the way over on the rubber against opposite-handed batters really do have the right idea. I tried running a couple regression analyses to see whether combining horizontal release point and arm angle resulted in an even stronger correlation, but that wasn’t the case. I suspect that horizontal release point just turns out to be a really good proxy for arm angle, and that it might even be factoring in the pitcher’s extension. I would never ever discourage a pitcher from scooching over on the mound all game, but I am sad to report that I no longer believe it could be that beneficial.

I’d also like to bring up one curiosity. When I ran the numbers this time around I checked both wOBA and xwOBA, just out of curiosity. Against opposite-handed batters, the numbers weren’t particularly notable. However, something interesting happened against same-handed batters.

For some reason, every single metric had a higher correlation to the expected wOBA than the actual wOBA. It moved arm angle past velocity! I don’t have a good explanation for this. Pitchers with lower arm angles perform better against same-handed batters, but their arm angle has a much stronger correlation to the expected wOBA of the batter than to the actual wOBA. Why could that be? When we talk about the difference between expected and actual stats, we’re just talking about the performance on batted balls, so we can ignore walks and strikeouts. If we limit things to balls in play, arm angle’s correlation to wOBAcon falls to .12, while its correlation to xwOBAcon rises to .26. If I had to guess, I’d say that it has something to do with spray angle. When the pitcher has the platoon advantage, batters tend to pull the ball less often, and pulled balls tend to outperform their expected stats by quite a bit. Maybe that variance is throwing things off, but I can’t draw any kind of straight line between that fact and the differences in this table. Someone smarter than I am is going to have to unpack this. As many FanGraphs readers match that description, please put any ideas you might have in the comments. I’d love to hear them.