Nick Anderson’s Strange Spin

On Wednesday night, Nick Anderson hung a curveball. Will Smith greeted it rudely:

Two batters later, Anderson faced Edwin Ríos with a chance to get out of the inning. As Eric Longenhagen noted on our Twitch stream at the time, Anderson hung another one:

Luckily for the Rays, Ríos didn’t quite time that one up. Anderson followed it up with another curve, which bounced, and he escaped the inning. Things could have gone much worse, however, and Eric and I mused that Anderson might want to take a look at what was causing his pitches to float in like that.

It seems pretty obvious that Anderson’s breaking ball, a biting snapdragon that seems to pack two inches of horizontal break into the last 10 feet of its homeward path, is at its best when it drops most. There’s only one problem with that theory: the data. Take a look at Anderson’s curve in 2019, broken up into quartiles based on vertical break:

Of note, I’m using the break calculations embedded in the Baseball Savant feed, which differ slightly from those that Pitch Info uses, but I’m more interested in the shape of the data than the exact magnitude of the break. This hardly makes any sense; the curveballs that hung up and actually rose very slightly relative to gravity produced the most swinging strikes and the highest whiff rate.

To put it bluntly, that’s not how I think of breaking balls (some might call Anderson’s curve a slider rather than a curve, though I hardly care which nomenclature you prefer). Downward break should be good. It moves the ball further away from the hitter’s bat as he swings, the exact vertical motion that seems to produce so many whiffs. No one ever watched a Pitching Ninja GIF and thought “I think that curve would be better if it just spun and spun instead of dipping.”

Naturally, I set to looking for what variable I’d missed. Was it velocity? I checked the velocity in each bucket:

Aha! Now we’re onto something. Maybe what we’re seeing isn’t break-related at all, but merely a reflection of the fact that faster pitches miss more bats. To test this, I split the data a new way. I broke Anderson’s 2019 curveballs down into quartiles by velocity, from lowest to highest:

Curiouser and curiouser. There doesn’t appear to be any particular relationship between speed and success. Break it down by 1 mph buckets, and it doesn’t get any clearer:

There’s no obvious connection or pattern here. Anderson’s breaking ball was excellent at most any velocity, but none seemed to unlock it. I won’t bore you with the particulars, but I dug through a few other possible causes: spin rate, active spin percentage, whether the same batter had just seen a fastball, and horizontal movement. Horizontal movement had the strongest correlation, which tells us something relevant: Anderson did best when he threw the pitch as a hard, sideways-breaking pitch, more of a traditional slider than a pure 12-6 drop:

Why is that? To be frank, I’m not really sure. Intuitively, the up-down breaker seems like it would pair better with his four-seam hammer, and missing bats just seems easier to do with pitches that break downward. Whatever the theory, though, that wasn’t borne out in practice by Anderson’s 2019, at least in any way I could find in the data.

With that 2019 data in hand, Anderson appears to have spent the offseason refining his arsenal. It’s hard to say anything conclusive given that he only threw 83 breaking balls in 2020, but the pitch resembled his “best” breaking balls from 2019: per Statcast, it dropped three inches less, to the point that his average fastball actually rose relative to spin-less motion. Pitch Info comes up with a different result — don’t quote me on this, but I believe that it includes a term for the drag force that air resistance exerts to counter gravity’s pull. Even there, however, his average breaking ball fell 1.7 inches less than the 2019 model:

I’ll be honest with you: I’m not exactly sure what this analysis means. Anderson’s breaking ball is an odd duck. It’s almost purely gyro spin, with an average active spin percentage of only 11% in 2019 and 7% this year. In other words, he’s throwing it almost like a football. Most sliders are thrown somewhere in between a football and a fastball, with closer to 25% spin efficiency; they break sideways far more than Anderson’s. Most curveballs are thrown with even more active spin and less gyro spin. In other words, Anderson is throwing a pitch with few true comparables.

With that in mind, it’s quite unclear what it means for one of Anderson’s breaking balls to “back up.” That nomenclature is usually used for sliders that a pitcher yanks, putting so much gyroscopic spin on them that they simply tumble like a bullet without breaking, the stereotypical “cement mixer.” That’s the way Anderson’s is supposed to work!

With the caveat that it’s only 83 pitches, here’s 2020 bucketed out into vertical break quartiles:

That nasty second quartile looks bad, but it’s only 21 pitches, not enough for a stable reading. There simply wasn’t enough season this season to say whether this plan has worked. One interesting takeaway, to my eyes, is that the horizontal/vertical tradeoff we saw last year is almost gone — he’s getting the same horizontal movement regardless of vertical movement. When he gets under the ball now, it’s simply moving the ball up during its flight without gaining any sideways run. One way to think about it is that in 2019, his “central” grip imparted slight topspin, which led to downward movement. When he “got under it” and gave it less topspin, that put more sidespin on the ball.

Think of a clock face: if the two-dimensional spin direction was normally at 8:00, getting under it, to say 9:00, increased sidespin and added horizontal movement. Now that he’s redone his grip, his normal breaking ball has a lot of sidespin already. When he gets under it now, though, that’s like going from 8:30 to 9:30 on the clock face; same sidespin, but backspin replacing topspin. On the cutter-slider-curve spectrum, the pitch has gone from in between a curve and a slider to in between a slider and a cutter, at least when he gets under it.

What does all of this mean? At this point, there’s a good deal of evidence that Anderson’s breaking ball does best when it’s moving mostly East/West. I don’t mean to say that the data is foolproof, merely that we have a decent chunk of data showing that batters miss that movement more often.

What’s less clear to me is whether that increased efficiency when he throws it “right” makes up for what happens when he gets under the pitches. In 2019, Anderson’s normal breaking ball was an excellent pitch. When he got over it, it bounced, out of harm’s way but with almost no chance of fooling the batter. Given that he often uses the pitch when he’s hunting strikeouts and has a ball or two to spare, that’s no big deal. When he got under it, it got even better. It’s a good outcome grid; the two fail cases when his mechanics were off were hardly a problem.

In 2020, that’s not the case. The normal breaking ball is now better, at least as far as whiffs go. When he gets over the top of the pitch, that’s fine too — that’s simply his normal breaking ball from last year, which we already know is an acceptable outcome. When he gets under it, though, I’m less sure. Now the break doesn’t move the ball out of the strike zone low, like last year’s pitch when he got over the top of it. Instead, it sits up without extra horizontal movement, giving hitters a crack at it.

I’m quite confident that I’m using the wrong words to describe all of this. Clock faces? Central pitches and fail cases? That’s me trying to make sense of Anderson’s new approach, but there’s no way he’s talking about it in the same clunky way I am. Whatever the linguistic faults, though, I think the analysis is really interesting.

Nick Anderson was one of the very best relievers in baseball last year, and he’s also been one of the very best relievers in baseball this year, playoffs aside. He’s doing it with a breaking ball that looks, to many observers, like a mistake pitch, but he’s doing it with intent. It’s a lovely little puzzle of a pitch, and I’m sure I haven’t yet cracked the code.