The Strike Zone Is Imperfect, but Mostly Unchanged

The strike zone doesn’t exist. Not physically, at least; it’s a rough boundary that varies based on how each umpire looks at it and how each batter stands. Catchers influence the shape, too; smooth hands can turn balls to called strikes, while cross-ups tend to do the opposite.

This year, the zone seems particularly amorphous — maybe it’s just my imagination, but I feel like I can’t turn on a broadcast without hearing about an inconsistent zone. Of course, hearing isn’t believing, and there are botched calls every year. Just because there have been some memorable ones this year doesn’t necessarily mean the overall rate of missed calls has changed. Let’s find out if it has, or if it’s merely imaginations running wild with the backdrop of fan noise.

For a rough idea of ball/strike accuracy, I went to Statcast data. For every pitch, Statcast records a top and bottom of the strike zone, as well as where the pitch crossed the plate. Armed with that data as well as some constants like the size of a baseball and the width of home plate, I measured how far out of (or into) the strike zone each pitch of the 2021 season was when it crossed the plate.

This data isn’t perfect. The top and bottom of the strike zone are approximated, and the plate isn’t a two-dimensional object, despite the fact that our data on it is represented that way. We aren’t considering framing. But we have previous years of the same data, which is great news. We can use the previous years to form a baseline, then see if this year’s data represents a meaningful change. And because we have a huge chunk of data, we can at least hope that framing comes out in the wash.

Let’s start with pitches outside the strike zone being called strikes. This year so far, a whopping 42% of pitches that missed the zone by half an inch or less have been called strikes. In the interest of getting an unbiased sample, I excluded skewed counts — 0-2, 2-0, 3-0, and 3-1 — so these aren’t gift strikes on unimportant pitches; they’re simply calls that turned pitches that were outside the strike zone into called strikes.

Disaster, right? That’s 42% of batters wrongfully denied their rightful ball. Umpiring is falling apart! Well, in the same counts last year, pitches that missed the plate by less than half of an inch were called a strike 43.5% of the time. Same same, which means we need to keep looking.

Those are pitcher’s pitches; batters understand that they’ll be victimized by those from time to time. What about when pitchers miss by slightly more? In 2021, pitches that miss the outside edge of the zone by between 0.5 and 1.5 inches have been called strikes 33.2% of the time. Aha! Those dastardly umpires. Except… in 2020, those pitches were called strikes 34.8% of the time.

Between 1.5 and 2.5 inches out of the zone, 17.4% of taken pitches have gone for called strikes this year. Last year, that number was 18.6%. Again, not much difference. Between 2.5 and 3.5 inches, umpires are calling strikes 7.4% of the time this year. Last year, they called those pitches strikes 8% of the time. We get it, we get it; pitches that just miss the zone are being called strikes at standard rates this year. Here it is in table form:

Called Strike Rate, Out of Zone Pitches
Distance from Edge 2020 2021
0-.5″ 43.5% 42.0%
.5-1.5″ 34.8% 33.2%
1.5-2.5″ 18.6% 17.4%
2.5-3.5″ 8.0% 7.4%

Okay, we have our first finding: on pitches that are off the plate but close, umpires are calling strikes at roughly similar rates compared to last year. How about pitches that miss by more than that? They basically don’t get called strikes, but the rates are similar from year to year. Though I’m not willing to take a strong stance given the limitations of a two-dimensional strike zone with moving top and bottom, I think it’s fair to say there’s no evidence of a change here.

What about pitches in the zone being called balls? That’s a slightly trickier calculation, but I used the same rough idea. Take the pitch’s location, add in the width of the baseball, then find the nearest edge of the strike zone — the distance to that edge is how far into the strike zone the pitch was. Pitches that just clipped the zone — between 0 and 0.5 inches from the edge — have been called strikes 61.2% of the time this year. Last year, they were called strikes 62.5% of the time.

Get further into the zone, and umpires call strikes more frequently — pitches that were in the zone by between 0.5 and 1.5 inches have been called strikes 75.4% of the time this year, as compared to 76% last year. Past that, it’s less automatic than you’d think, though still fairly automatic, to the tune of an 86.8% called strike rate on pitches in the zone by between 1.5 and 2.5 inches. That compares to 88% in 2020. Here’s the same table as above, only for pitches in the zone:

Called Strike Rate, In Zone Pitches
Distance to Edge 2020 2021
0-.5″ 62.5% 61.2%
.5-1.5″ 76.0% 75.4%
1.5-2.5″ 88.0% 86.8%
2.5-3.5″ 93.2% 91.5%

Umpires are calling slightly fewer strikes in every case this year; that’s the only clear trend in this data. If your baseline is what umpires have done before now, they’re doing roughly the same thing this year. That doesn’t mean that they didn’t blow a call when your favorite pitcher needed the strike, or punch out your team’s top slugger on a pitch three inches outside. It just means they were doing that last year, too.

You could imagine a different grid, though. Here’s the theoretically optimal grid, if you care most about accuracy:

Theoretically Optimal CS%, In Zone Pitches
Distance to Edge 2020 2021
0-.5″ 100% 100%
.5-1.5″ 100% 100%
1.5-2.5″ 100% 100%
2.5-3.5″ 100% 100%

Bring on the robots! They’ll get this stuff right. I have two counterpoints to that — not disagreements, per se, but arguments. I think there’s value in uncertainty — I like the fact that pitches that clip the edge are strikes 60% of the time, while pitches an inch in are strikes 75% of the time. That probabilistic nature makes sense to me. Bright lines are weird! A scale of called-strike-iness is how I imagine the strike zone working, and I like that it works that way in real life.

A related concern: when you’re dealing in absolutes, slight measurement errors are important, and no system is perfect. For example, take a look at this called strike:

That missed the zone; no arguments about that. But per the data, that pitch was 4.5 inches low, the single most egregious called strike in the so-called “shadow zone” this year. It certainly doesn’t look that way to me on the video.

Part of the reason for that: the bottom of Jazz Chisholm’s strike zone gets recorded independently on every pitch, and it was 1.5 inches higher on this particular pitch than its year-long average. The bottom of his zone has varied by 6.72 inches from highest recording to lowest recording this year in 211 pitches. I’m not privy to the calculations that go into determining the top and bottom of a player’s strike zone on each pitch, but that’s a wide range, and might result in pitches that annoy us just as much as the current “missed” calls.

I doubt this article will change your feeling about umpires one way or another. If you wanted the robots already, you probably still do. If you liked the uncertainty and tradition of the current system before this year, you probably feel the same. Umpires don’t appear to be getting worse at calling balls and strikes, no matter what announcers’ eyes are telling them. The question, instead, is whether their current level of precision is the kind of strike zone you want.





Ben is a writer at FanGraphs. He can be found on Twitter @_Ben_Clemens.

newest oldest most voted
The Stranger
Member
Member

“The bottom of his zone has varied by 6.72 inches from highest recording to lowest recording this year in 211 pitches.”

This is a thing I did not know. Unless the robot umps are programmed to use a fixed value for each batter, a six-inch zone of “maybe a strike” seems like it’s not really any better than human umps.

HappyFunBall
Member
Member
HappyFunBall

It seems like a lot. But if Chisholm varies his batting stance significantly between PAs (or between pitches) it may even be right

Anon21
Member
Anon21

I’ve never given this much thought before, but is that how we want the zone to be called? Like a guy can kind of scrunch up to buy a ball?

Sæder
Member
Member
Sæder

Funny. I read the wikipedia article on the strike zone the other day (FWIW, it could be inaccurate, but I’m lazy) and it quoted the rules as saying that the zone is defined by location of “letters” and knees of the batter in his stance. So already, explicitly, I believe that this is how it works. Scrunching up likely does have an effect on the zone (though I’m guessing more for the top of the zone than the bottom since your knees can’t move that much since your shin bone is fixed?)

PS: the history of the strike zone in the 19th century is fascinating! I didn’t know that batters used to tell the pitcher the height at which to throw the ball! Nor did I know that a called “ball” is short for “no ball”. To be stupid, next time someone asks me if I think that was a ball, I’m gonna reply, “no ball” 🙂

dl80
Member
Member
dl80

Good point. This would be very easy to do, I think: have all the uniforms have some kind of sensor/reflector sewn in at the letters and the approximate knees. The robot umps then use them to determine the zone for each batter. If a batter wants to scrunch way up, so be it.

Pwn Shop
Member
Member
Pwn Shop

You could easily crouch such that your strike zone has zero size. Of course, you could do it today but i think the human ump might still wring you up.

Sæder
Member
Member
Sæder

This made me wonder about the uniformity of the position of “the letters” on jerseys. Or how a jersey must fit so that an ump knows where the top of the zone should be.

Also, do the players then wear a sensor to delineate the zone so that the robot ump knows where it is? How does that sensor not get shifted around accidentally, or even intentionally?

Getting flashbacks to my HS JV baseball days where our coach would scold us harshly if we ever said anything to umps about the balls and strikes. haha

Spahn_and_Sain
Member
Spahn_and_Sain

It won’t be a sensor, the Hawkeye system is camera based computer vision. They’ll use pose estimation.