“’Cause it’s gonna be the future soon,
And I won’t always be this way,
When the things that make me weak and strange get engineered away.”
--- Jonathan Coulton, “The Future Soon”
The Invisible Box
Every baseball season, calls get louder for Major League Baseball to implement an electronic strike zone.
We see it on our TV screens: that little box showing exactly where the imaginary strike zone is supposed to be. We see pitch locations pop up instantly on our phones with MLB’s own AtBat app. All these are visible to millions of viewers thousands of miles away, but not to the home plate umpire standing just a few feet behind home plate.
The whole world knows immediately if a pitch was a ball or a strike. Why doesn’t Major League Baseball let the umpire, the one person who absolutely needs to know, in on the secret?
And when the postseason comes around, more rides on every pitch. With every missed call, Amazon starts selling out of pitchforks and torches.
In Game 4 of the 2019 ALCS between the New York Yankees and the Houston Astros, home plate umpire Dan Bellino called this a strike on Jose Altuve:
And this a ball to Aaron Judge:
How much does a missed ball/strike call really matter anyways, though?
The value difference between a ball and a strike has been quantified. Dan Brooks and Harry Pavlidis of Baseball Prospectus found the difference between getting a strike call and a ball call to be roughly 0.14 runs. That doesn’t sound like much, but a few calls flipped one way or the other and it can add up.
A 3-1 count instead of a 2-2 count dramatically shifts the favor of an at-bat towards the hitter and away from the pitcher. In a critical at-bat, this could potentially swing the outcome of an entire game. 0.14 runs is only the weighted context-neutral average. Depending on the situation, the difference between a single pitch being called a ball or a strike can be up to 0.59 runs. Dan Meyer at the Hardball Times took Brooks’ and Pavlidis’ work one step further and found that the quality of the hitter at the plate also factored in as well.
The tide can turn on the smallest of differences in a baseball game. Chalking up missed ball/strikes to “acceptable losses” can take the game out of the players’ hands and into the umpire’s instead. It is important to get the call right.
Dan Bellino is by no means a bad umpire. Umpires who call Major League Baseball games are in that position for a reason. Just as players who make it the Show are the best baseball players in the world, MLB umpires are the best umpires the sport has to offer.
MLB umpires are constantly getting better too. Boston University studied every pitch thrown from 2008 to 2018, and the number of bad ball-strike calls decreased every single year from 16.36% in 2008 to 9.21% in 2018. But they’re human. They won’t ever reach 0.0%, and 9.21% is still 34,294 missed calls over the course of the 2018 season.
Still in Beta
So why isn’t the electronic strike zone here yet?
Well first, that little box we see on our TV screens isn’t really the strike zone, and that small white dot showing us where the pitch crossed the plate isn’t necessarily accurate.
Here are Brett Gardner (left) and Aaron Judge (right) batting in Game 2 of the 2019 ALDS between the Minnesota Twins and the New York Yankees. The FS1 broadcast has kindly provided the faint white outline of the strike zone. I’ve added in the yellow lines.
FS1 is using the same strike zone box for 6’7” Aaron Judge as they are for 5’11” Brett Gardner. The top of Judge’s strike zone doesn’t even reach his jersey. It’s aligned at the level of Judge’s belt. As far as FS1 is concerned, the strike zone is Judge’s pants.
How can FS1 be that off? Consider that FS1’s primary objective is not to enforce the rules or preserve the integrity of the game. Their objective is to create the best viewing experience for the viewer. That might mean getting strike zone boxes up on the screen more quickly by forgoing redrawing the box for each hitter. That might also mean sacrificing accuracy of pitch location by cutting down on the number of calculations, if it means that little white dot can pop up on your TV a fraction of a second earlier.
There are other barriers. In 2018, Baseball Prospectus outlined the numerous obstacles that MLB and current technology still face in creating a reliable electronic strike zone. (This is a highly recommended read.) The accuracy, though continually improving, is nowhere close to how accurate the public thinks it is. This doesn’t just apply to the TV broadcast, but to MLB’s Statcast pitch tracking system as well.
There are also logistic issues that require addressing when the machine breaks. (And don’t kid yourself, every machine breaks.)
Revisiting the topic two weeks ago, Rob Arthur of Baseball Prospectus again concluded that MLB just isn’t ready for robot umpires yet. Progress is being made, but the system still needs a lot more debugging. “A botched rollout of robo-umps would make the scandal surrounding the ball look trivial by comparison and probably kill the electronic strike zone cause for years or decades,” concludes Arthur.
Despite these still present obstacles to an electronic strike zone, robo-umps are coming. At the behest of MLB, the independent Atlantic League implemented an electronic strike zone for their 2019 season. Although there were a few hiccups, it went well enough that MLB has stated they will continue to refine the process in 2020 in the minor leagues.
It is no longer a question of if MLB will adopt an electronic strike zone, it is simply a matter of when.
The robot invasion is inevitable.
The End of Framing
Catcher framing is the art of receiving a pitch in a manner that persuades the home plate umpire to call strikes as strikes, and even borderline balls as strikes. It is a real skill. Catchers like Austin Hedges, Tyler Flowers and Max Stassi are routinely among the league leaders in framing, as measured by Baseball Prospectus’ CSAA (Called Strikes Above Average).
It is also an incredibly impactful skill. Earlier this year, I examined the difference in runs above average between catchers at the 90th percentile of a particular skill and the 10th percentile. The magnitude of impact of having a good framing catcher as opposed to a bad framing catcher was on the same order of magnitude as the impact of having a good hitting catcher as opposed to a bad hitting one.
For as much attention as people pay to a catcher’s OPS and wRC+, a catcher’s CSAA and framing skill is nearly as important.
Catcher is considered the most defensive-oriented position in the game, apart from the pitcher. The catcher’s defensive skills: framing, throwing and blocking, constitute nearly half (45.8%) the above pie graph. Most of that half (31.3%) comes from framing.
Framing is a peculiar skill. The strike zone is very clearly defined in the MLB Rulebook:
Rule 2.00: The Strike Zone
The STRIKE ZONE is that area over home plate the upper limit of which is a horizontal line at the midpoint between the top of the shoulders and the top of the uniform pants, and the lower level is a line at the hollow beneath the kneecap. The Strike Zone shall be determined from the batter’s stance as the batter is prepared to swing at a pitched ball.
By framing, catchers are attempting to alter how the home plate umpire enforces the strike zone so that it favors their team. That’s not playing within the rules; that’s playing with the rules. Ideally, that should not be allowed.
When electronic strike zones are implemented, Framing’s 31.3% portion of the pie will drop to zero. Short of carrying around a laptop and an USB cord, a catcher can’t influence a robot. Balls and strikes will be called independent of how the catcher chooses to position his body or mitt.
MLB catchers will go the way of Aaron Brothers stores: out of the framing business.
With the elimination of framing, the new distribution of the relative impacts of catcher skills looks like this:
Catcher is now a more offensive position.
Catching the Tail of the Trend
Which catchers suffer? Which catchers benefit?
Catchers who excel at framing will naturally be the ones who will most miss human-called balls and strikes. Last week, I published Catcher Skill Spectrum Scores for the 2019 season for catchers who caught > 500 innings and had > 250 PA. If the weighting of each skill is changed so that framing skill no longer carries any weight, these five catchers see the largest decreases in their overall score:
Conversely, these five catchers, for whom framing was one of their weaker skills, would see an increase in their overall score:
But although some catchers are now viewed better and others worse, it’s not a wash. Getting rid of framing makes the entire pool of catchers more homogenous. It would be as if all NFL quarterbacks were somehow magically made to have the same arm strength, or if all NBA centers were somehow made to have identical free throw percentage rates.
By eliminating a skill that a player could be better than their peers at, you limit the ways for a player to stand out at their position and be more valuable.
Philadelphia Phillies catcher J.T. Realmuto excelled in all five skills in 2019. Though he was quite good at framing, it rated as the weakest of his five skills.
By eliminating framing, his overall score increases from 87 to 94, but this is relative to other catchers. His overall value as a position player is hurt. Realmuto accumulated 37.7 Catcher runs above average (Batting Runs + Baserunning Runs + Framing Runs + Throwing Runs + Blocking Runs) in 2019. Without his 10.5 Framing runs, he would have only accumulated 27.2 Catcher Runs. That is still a lot, but it is still less than he would have had when he was able to use his framing ability to stand out further from the pack.
A well-implemented electronic strike zone may be good for the sport, but it does diminish the role of the catcher by taking away potential defensive value.
After catcher, shortstop is next in traditionally perceived defensive value on the defensive spectrum. Earlier this year, I compared the relative skill differentials between 90th and 10th percentile catchers to that of shortstops. Whereas defense comprised nearly half the pie for catchers, it constituted just a little more than a fifth of the pie for shortstops.
After eliminating framing, compare the two pie charts:
They are nearly identical.
Without framing, a catcher’s ability to block becomes the catcher’s most impactful defensive skill, but it is nowhere near as impactful as framing was.
This brings us to another rule for which MLB had the Atlantic League serve as guinea pig.
“Stealing” First Base
This past year, MLB asked the Atlantic League to implement a curious rule of allowing a batter to try to advance to first base on any pitch that gets past the catcher, whether it be a wild pitch or a passed ball. Although it is informally known as “stealing first base,” it is really more akin to the scenario of reaching first base safely on a dropped third strike. The difference is that now it doesn’t have to be strike three to happen.
It seemed a bizarre experimental rule, deviating more from traditional baseball rules than any of the other experimental rules. Joe Sheehan in his newsletter proposed one possible reason this rule was included: by creating more blocking chances, you add some defensive value back to the catcher that was lost with the elimination of framing:
One of the arguments against automated ball-and-strike calling is that it would eliminate the need for catchers to provide value defensively. I submit that the “stealing first” rule would balance that, adding to the value of catchers who can receive and block and throw, even as the value of “framing” is eliminated.
How much defensive value is added by instituting the “stealing first” rule? As it is, blocking is already an important skill. With runners on base, a ball that gets past the catcher potentially means an extra base for every runner. Even without runners on base, blocking is important with 2 strikes, so that the batter doesn’t reach first on an uncaught third strike.
From 2016 thru 2019, catchers saw roughly 44 blocking chances in 9 innings of play.
Blocking chances occur whenever a pitch does not contact the bat and there is a possibility of runners or the batter advancing if the pitch gets by the catcher.
With the bases empty and an 0-0 count, there is no blocking chance. Even if the ball goes to the backstop, there are no runners or batters who could advance.
With the current rules:
- Any pitch with a runner on base where the bat does not make contact (ball, called strike, swinging strike) would be a blocking chance. There were 228,759 of these pitches in 2019 per Baseball Savant. (This is less so on ball 4’s, but theoretically a runner might be able to advance more than one base if the ball gets far enough away.)
- Any strike three pitch with the bases empty is a blocking chance, as the batter could run to first on a dropped third strike. There were 23,947 of these pitches in 2019.
- Together, that is 252,706 pitches where there was some import to the catcher making sure the pitch did not get past him.
With the ability to run to first base on any pitch that gets past the catcher, more blocking chances are added:
- Any pitch with the bases empty that is a ball 1, 2 or 3 is now a blocking chance. (Ball 4, the runner would advance anyways.) There were 144,997 of these pitches in 2019.
- Any pitch with the bases empty that is a non-fouled strike 1 or 2 (looking or swinging) is now a blocking chance. There were 97,864 of these pitches in 2019.
- Together this is an additional 242,861 pitches where blocking matters.
The number of pitches in 2019 where blocking would matter increases from 252,706 to 495,567. This is an increase of 96%, which nearly doubles the number of blocking chances.
With the number of blocking chances increased by 96%, we can estimate the number of Blocking Runs above average that a 90th percentile catcher would have over a 10th percentile catcher would increase by 96% as well. This is not precisely true, because the run expectancy of an extra base varies depending on the count and to which base a runner advances, and all the additional blocking chances are with the bases empty. However, the consequence of a passed ball also increase in blocking chances with runners on, because the batter can now advance along with the runners.
For our purposes, we will estimate that a 96% increase in the number of blocking chances results in a roughly equivalent 96% increase in blocking runs. In that case, the difference between a 90th percentile blocking catcher and a 10th percentile blocking catcher would be 15.6 runs instead of 7.9 runs. The Relative Skill Differential pie graph would look as follows (with framing eliminated):
Some defensive value is restored (Now 32.1% of the pie instead of 21.1%), but the defensive portion of the pie is still well below what it was before implementation of an electronic strike zone (45.8%).
This rule seems unlikely to make it to the majors. While the electronic strike zone has been green-lighted for use in the minor leagues in 2020, there has been no such similar announcement for the “stealing first” rule.
Even without the “stealing first” rule, the robot invasion is coming. And it will be good. Letting robots take over the world; what could go wrong?