J.T. Realmuto as Bizzaro Mike Piazza

Catcher defense is all the rage, at least for some. For those of us who care deeply about catcher defense, this year has brought us glimpses of Statcast and Baseball Prospectus’ recently gathering of enough data to extend their pitch framing, blocking, and throwing metrics back as far as 1950 in some cases. We’ve always known about catcher defense, but our ability to measure and understand it has improved greatly in the last five years, and more advances are likely on the horizon.

This ability to measure something affects our perceptions not just of that thing, but of all other things related to that thing. As Ben Lindbergh recently noted, Mike Piazza was known as a horrible throwing catcher during his career and that colored his entire defensive perception. Yet our modern metrics look back favorably upon his blocking and framing. It’s important to note that we didn’t just learn about blocking and framing in 2015. It’s not as if baseball fans couldn’t conceive of their value in 1997, it’s just that the only thing we had rudimentary numbers for at the time was his arm. We could measure that and see it most clearly relative to the other skills, so it did the heavy lifting in our minds.

These days, we have much more data. 2016 will be the 10th season for which we’ll have some PITCHf/x data, and it will be the ninth season for which we have data on virtually every pitch. Even if you’re not totally sold on some of BP’s efforts to take framing back to the 1980s, we have plenty of PITCHf/x data to satisfy your recent curiosities. Maybe you don’t agree with the exact methodology of the various framing runs metrics, but they are a vast improvement over trying to eyeball framing value, and these metrics have matured in the last few years, past things like caught stealing rate, rSB (runs saved), and RPP (blocking runs) to much more complete metrics. The data has been illuminating, and for the past few seasons, we have judged catchers primarily based on their framing because that’s the part of their game we could best measure. It’s natural, but it’s partially wrong.

Catching is about throwing, framing, and blocking. (There’s also game-calling/management, but we’ll leave ourselves to consider only the physical skills for now.) Those three skills matter, and they don’t matter proportionally based on our ability to measure them. The importance of framing does not increase because we can measure it; our ability to measure just allows us to make decisions based on a better representation of the truth.

Imagine you had three envelopes, each perfectly describing a catcher’s abilities along the three skill dimensions. Now imagine I tell you that you can open one envelope and have to decide whether you want to sign him based on that data (assume you know his offensive traits as you normally would). Whichever envelope you choose, the information in the other two remains the same, you just can’t observe it. But the information you do see will probably color your impressions of the data you can’t see.

If you see that a catcher is a good framer, maybe you think that means he’s a good blocker. If he’s a bad thrower, maybe you think that means he can’t frame. It’s not that you are sure of any of those inferences, it’s that your brain is trying to fill in the gaps. But as we recently learned with Piazza, this isn’t always true.

We can see this in Marlins catcher J.T. Realmuto. If you’re familiar with framing numbers, you know that Realmuto graded out very poorly in 2015. Baseball Prospectus says he cost the Marlins -15.7 runs through poor framing last year, and as a sanity check on those numbers, StatCorner says it was -14.6 runs. In other words, the best public framing data suggests he performed very poorly in this regard in his first full season.

I don’t mean to suggest that Realmuto is hopeless, merely that so far he has failed to deliver extra strikes for his pitchers. Interestingly, in fact, BP’s newly minted minor league framing numbers suggest that Realmuto was a talented framer in the minor leagues. Their minor league data isn’t based on PITCHf/x, so it is possible that’s a statistical anomaly, but I discussed Realmuto’s numbers with Jonathan Judge at BP who confirmed that if you apply the “minor league method” to Realmuto’s 2015 MLB numbers, he grades out poorly. In other words, the different method does not appear to explain the shift from good framer to bad.

At this point, I can’t litigate what changed for Realmuto from 2013-14 to 2015, but I can say that based on what we know about framing at this point in time, he seems to have cost his team about 1.5 wins based on his lack of aptitude in 2015. Is that perfectly accurate? Who’s to say, but the size of the impact is large enough that it’s hard to question that there is some signal in the noise.

These powerfully negative framing marks would seem to stick Realmuto with a negative defensive profile. He’s performed like an average hitting catcher, without substantial room for offensive growth according to erstwhile prospect analyst Kiley McDaniel, so if his glove is really as bad as the framing numbers make it look, he’s not a major league regular.

But there is another side to this story. BP’s metrics believe his slightly below average blocking and above average throwing come out in the wash, but the magnitude of both come in far below the magnitude of his framing. In other words, he might have some other skills, but they don’t seem to make up for his trouble managing the zone.

Yet we do have the ability to go beyond BP’s metrics when it comes to throwing. While their methodology is a good one, it’s one based on averaging out the effects of other participants. It’s not truly about the performance of the player once you go beyond framing. Given the data we have publicly, it’s a fine way to go, but with some Statcast numbers we can dig in further.

As you might know if you’ve picked up a copy of the Hardball Times Annual, Statcast tracked 1,784 pop times in 2015 and J.T. Realmuto had the best average time of the entire group. This is true even without applying a minimum number of throws to eliminate players who only made three or four tosses to second.

Realmuto’s average pop time was 1.867 seconds (the league average was 1.975). To put that in perspective, Realmuto’s average pop time was faster than 84% of all throws to second base. He had five of the best 11 throws, seven of the best 25, 12 of the best 45, and 16 of the best 93 in 2015. He only made four throws slower than 1.941 seconds.

To give you an idea, this is his sixth best throw from 2015 (18th overall), coming in at 1.783 seconds:

Using objective data, it’s clear that Realmuto is a superior thrower even if his caught stealing rate doesn’t jump off the page. But to know this about Realmuto, you need Statcast or you need to watch the Marlins regularly with a keen eye for catching. Otherwise, there’s nothing out there to show an interested, non-Marlins fan how talented he is at this facet of the game. It’s not hidden information, but it’s awfully close.

This is interesting because what we have before us are two independent statistics that tell opposite stories about Realmuto’s defense. He is simultaneously one of the worst framers and arguably the best thrower. Because the framing data is complete and public, our general estimates of his value are driven largely by the former. He is merely an average hitting catcher who can’t frame, unless you happen to have access to one of the envelopes that tells you Realmuto is the most punishing thrower in the league.

I won’t attempt to translate his pop times to runs saved, expected runs saved, or something of that nature. It’s very likely that framing matters more than stolen base prevention when it comes to catcher defense, so it’s not as if Realmuto has completely atoned for his negative framing. But on the other side of the coin, the margins in baseball are small and if you put his great throwing on one side of the scale with the knowledge that he seemed to be a much better framer in the minors, you might have a different outlook on his overall defensive ability than if you only looked at his 2015 MLB framing.

Just because we can’t measure something doesn’t mean we should assume it will line up with the other things we know that are related. Piazza and Realmuto are good evidence of that. As a final test, I lined up 2015 framing runs (per 7,000 chances) with average pop time (minimum 20 tracked throws). It might seem like a mismatch of sample sizes, but pop time seemed to stabilize quickly based on my work with them for the Annual. Does Realmuto stand alone as a catcher whose throwing and framing point in opposite directions?

Generally, it’s Realmuto and Tyler Flowers for opposite reasons. Realmuto is the only great thrower with horrible framing numbers and Flowers is the only great framer with really bad pop times. There isn’t much of a relationship overall, but these two do stand out for opposite reasons when looking at the data.

This isn’t to say that framing isn’t the most important aspect of a catcher’s game, but rather that the parts of the game we can’t measure very well aren’t necessarily like the parts of the game that we can measure. This is a lesson specific to Realmuto, but one that also seems to apply broadly. If you don’t have good measurements, you don’t want to throw up your hands and do nothing, but you have to be willing to accept the possibility that the data you can’t see contradicts the story being told by the data you can see. After all, in 2015, the worst framer was the best thrower. There’s no logical reason to expect that those skills would correlate, but J.T. Realmuto is a good reminder not to take anything for granted as we step out into the second and third generation of advanced metrics.