# On Framing and Pitching in the Zone

One of the most interesting fields of study in baseball over the last few years has been that of pitch-framing, or pitch-receiving, or pitch-stealing, or whatever you want to call it. This is the stuff that’s made Jose Molina nerd-famous, and it’s drawing more attention with every passing month. Framing has been discussed on ESPN. It’s been discussed on MLB Network. It’s been the subject of countless player interviews, and what’s been revealed is that a great amount of thought and technique goes into how a catcher catches a pitch. Catchers don’t just catch the baseball. They catch the baseball with a purpose.

Research has uncovered a few outliers, like Molina and Jonathan Lucroy and, say, Jesus Montero and Ryan Doumit in the other direction. It’s interesting these guys can be given such different strike zones, since the strike zone is supposed to be consistent for everybody. And it’s interesting that, as much as people come up with run values in the dozens, it’s hard to identify the actual effect. For example, Rays pitchers this year have allowed a higher OPS throwing to Molina than when throwing to Jose Lobaton, the other guy. Last year, Molina again had the worst numbers. It reminds me too much of Catcher ERA for my tastes, but you’d think you’d see something. Instead, you see little. Where is the value going?

In this post, I’ll ask more than I answer, and this should be a jumping-off point for other, better researchers. It seems to me the effect of pitch-receiving is often overstated, and I’ve been searching for explanations why. I do have one theory, and before we get to that, another explanation of the same old familiar home-brewed stat. The stat is named Diff/1000, and it’s the difference between actual strikes and expected strikes per 1,000 called pitches, as derived from plate-discipline data available on FanGraphs. A positive Diff/1000 means a player or team got more strikes than expected. A negative Diff/1000 means the opposite. Whenever I calculate Diff/1000, now, I adjust it to set the league average at zero. The league average is always below zero, by a decent amount.

I went back and collected all team pitching data from between 2008 and 2013, and I calculated Diff/1000. Then I plotted Zone% against that stat, yielding the following graph:

As teams have been better at receiving — that is, as they’ve generated a greater rate of extra strikes — they’ve thrown a lower rate of pitches in the strike zone. The relationship is remarkably strong — given what we’re dealing with — and the bottom fifth in Diff/1000 have an average Zone% of 51.3%. The upper fifth in Diff/1000 have an average Zone% of 48.4%. Broken into quintiles of 36 teams:

• Quintile 1: 48.4% Zone%
• Quintile 2: 49.4%
• Quintile 3: 49.8%
• Quintile 4: 51.0%
• Quintile 5: 51.3%

I think there are three possibilities: The first is the most obvious. If a team knows it can get strikes off the edge, it’ll pitch to the edge more often. A pitcher will be more likely to miss the strike zone because the catcher might set up on the fringes, knowing he’s capable of turning some of those pitches into strikes. A strike on a pitch on the edge of the zone — or out of it — is a good strike, because those are difficult pitches to hit. Pitchers won’t stay in the zone if they don’t have to.

Another explanation goes in the reverse. Maybe pitchers have been getting extra calls because they’ve been pitching to spots out of the zone. Maybe instead of framing earning pitches, pitches have “earned” framing. But then, it’s been demonstrated that there are different receiving techniques, and some are better than others. So. The first two explanations are kind of related to one another.

And then there’s the possibility that this is just circular. Zone% is used in the calculation to derive Diff/1000, so maybe that’s just what I’m picking up. I am posting this knowing I might get exposed as an idiot, but at least then I could learn something from being called out. I noted before that I’m mostly asking instead of answering. I’d love to know if I’m screwing something up, because I’m interested in this though I’m not all that intellectually powerful.

Let’s move forward. Teams that have been better at receiving have thrown a lower rate of pitches in the zone. Therefore, even though they’re getting extra strikes on pitches off the edge, that should just be counter-balancing the overall strike rate. Let’s plot team strike rate against Diff/1000:

Here, we see almost no relationship. There’s a slight positive slope, but here are those same quintiles, from best to worst Diff/1000:

• Quintile 1: 63.3% Strike%
• Quintile 2: 63.0%
• Quintile 3: 62.7%
• Quintile 4: 63.0%
• Quintile 5: 62.8%

In terms of actually getting strikes, the best receiving teams have barely been better than the worst receiving teams. This is because the worst receiving teams have thrown more pitches in the zone than the best receiving teams. This seems like it could be where a lot of the value is going. Good receivers have generated good numbers of extra strikes, based on the pitches they’ve caught, but they’ve caught a lot of pitches out of the zone so the overall strike rate still looks fairly normal. Which could explain why we don’t observe that much of a boost.

Of course, a strike that’s called is better than a strike that isn’t called because a non-called strike could be a ball in play and balls in play can be bad. While there’s little difference in overall strike rate, there’s a greater difference in called strike rate and that’s why it would be better to be good at receiving than bad. But I’m still skeptical that we’re dealing with differences of dozens upon dozens of runs. It looks like a chunk of the added value is lost from throwing more pitches out of the zone, but I’m open to other ideas. This is not the last word on anything.

As a quick case study, Derek Lowe got a ton of extra strikes from 2008 to 2011, when Russell Martin and Brian McCann were his primary catchers. In 2012, those extra strikes went away, as Lowe joined the Indians. Between 2011 and 2012, Lowe’s Zone% increased from 37% to 47%. You can see some differences in the following heat maps, in which 2012 is on top and 2011 is below:

Lowe couldn’t pitch off the edge as much, which cost him out-of-zone strikes. But he just pitched in the zone more often, such that his overall strike rate barely changed. Of course, Lowe was also terrible in 2012, so this could be a case where Lowe perhaps depended on the quality receiving. Maybe, in this case, framing was a huge help. As a different case study, there’s Kyle Lohse, who’s joined this year’s Brewers, albeit after pitching to Yadier Molina. The Brewers lead baseball this year in Diff/1000, and Lohse’s Zone% is down from 51% to 49%. His strike rate is exactly the same. And so on and so forth.

I hope that others keep writing and keep researching, because this stuff is interesting. Good and bad pitch-receiving, unquestionably, makes a difference. I just badly want to know how much of a difference it makes. I’m unsatisfied that we have the real answer right now.

Print This Post

Jeff made Lookout Landing a thing, but he does not still write there about the Mariners. He does write here, sometimes about the Mariners, but usually not.

Guest
ecocd
3 years 3 months ago

I read almost everything you put up anyway, Jeff, but I never miss a pitch-framing article. Whenever you feel the itch, don’t worry about beating the topic into the ground. The Lowe pitch charts are fantastic.

While you claim that you think pitch framing may be overstated, I would argue they’re incredibly valuable. My introduction to sabermetrics came through Moneyball (given that I’m on Fangraphs, I’ve progressed from there) and the one thing that stuck with me was the BA in different counts. It seems generally well known that a first strike gives the advantage to the pitcher, it’s the fact that BA change so dramatically that caught me off-guard. The difference between 1-2 and 2-1 is about .160. Getting an extra strike on 1-1 is incredibly impactful.

Guest
Andrew
3 years 3 months ago

Wouldn’t it be better to analyze OBP from a specific count, rather than simply BA? I’m sure the trend would continue (and might event be stronger), but I thought sabermetrics has taught us that BA might be the most overused stat in baseball (well, offensive stat, surely nobody will defend pitcher’s win/loss records as meaning anything).

Guest
Sam
3 years 3 months ago

Would it be possible to take the run values that Marchi or Fast originally published, add those to the team run totals and see if those figures were closely related to actual runs scored? If framing at the league/team level greatly overstates actual run scoring then it’s probably being overstated at the individual level.

Guest
Neil
3 years 3 months ago

The zone issue likely comes from the fact that you can’t frame pitches that were already in the zone, so if a team throughs more pitches in the zone, there is less of an opportunity to develop good Diff numbers.

Second, can you isolate by pitch type? In reading about the way the brain perceives speed I saw a big connection here. Umpires might have a tough time seeing the last few feet of a fastball, especially if it has a good deal of movement late in the process. When the pitch moves late, their brain has to fill in between the last place they saw the ball and the glove, therefore, giving more weight to where the pitch is caught. Sliders could work too, but less so with changeups and curves.

Guest
Neil
3 years 3 months ago

Throws*, whoops

Guest
Tim
3 years 3 months ago

Another explanation might be that teams with good framers are ahead in the count more often, and thus more likely to throw balls. Breaking this down by count would be useful.

Guest
Synovia
3 years 3 months ago

I agree with this. The fact that the overall ball/strike percentages come out the same doesn’t mean that there’s no difference in value. its entirely possible that these better framing catchers are putting their pitches in much better counts.

Guest
tz
3 years 3 months ago

True. It would be interesting to see if the Diff% varies by count, to see there are broad tendencies for umpires to give borderline calls to whoever is behind in the count.

Guest
Synovia
3 years 3 months ago

“True. It would be interesting to see if the Diff% varies by count, to see there are broad tendencies for umpires to give borderline calls to whoever is behind in the count.”

There’s another article on here that shows that umpires generally err on the side that keeps the at-bat going. Borderline 3-0 pitches are almost always strikes, borderline 0-2 pitches are almost always balls, etc.

Guest
Youthful Enthusiast
3 years 3 months ago

Another thought is that pitchers don’t have to “give-in” in hitter’s counts. If you’re able to throw a quality strike on the edge of the zone on 3-1 rather than a fastball down the middle, you’re going to do better than the averages would expect you to.

This is something we could test. Do the better pitch framers do perform better than expected in each count? Hitter’s counts? Pitcher’s counts?

Guest
Youthful Enthusiast
3 years 3 months ago

Something to consider is the “quality” of the strike. A pitch down the middle is much easier to hit than a pitch on the edge. If a pitcher can’t reliably get the calls on the edge, he has to throw more over the heart of the plate. More balls over the heart of the plate mean more more contact, harder contact, and hits.

When you talk about the strike rate and zone% balancing out, it kinda makes sense. If the pitcher can get strikes where the batter is less likely to make good contact, they’re gonna throw it there. Even though the strike rate isn’t going up, the “quality” of the strikes is increasing. If a pitcher could replace 10 strikes down the middle with 10 strikes right on the edge, what is the value in that? The fact that strikes are binary doesn’t reflect the continuous nature that moving a pitch outward from the strike zone displays.

Guest
Jeff
3 years 3 months ago

Another thing to consider is the intentions (conscious or not) of the umpire. I wonder if umpires subconsciously approach each game with some idea of what percentage of pitches are likely to be balls and are likely to be strikes. Over time, this rough balance gets reinforced. If a pitcher is extremely wild, the umpire adjusts this expectation. However, for pitches that are “close” (the ones likely to be most effected by framing) perhaps this is where this a priori expectation could come into play. It could be that pitchers that throw a lot of pitches in the zone simply aren’t going to get as many framed strikes because they have already hit the expected percentage of strikes. This is not to say that umpires are doing this intentionally, but umpires obviously are an important factor here.

I would also second the need to examine this by count. It seems like this is an area where it could really shine some light.

Guest
marc w
3 years 3 months ago

You’d think you’d find some difference in either swinging strikes or BABIP given that the guys with good framers throw more balls. You’d be giving back some of the run value, but that should be balanced by better outcomes on balls in play. Of course, all of that would show up, theoretically, in runs allowed.

Here’s something that is hopefully easier for someone: do pitchers throwing to great framers see more balls put in play on pitches out of the strikezone or is it just the same? The huge run values would seem to argue for the former, but the fact we don’t see huge CERA-ish differences argues for the latter.

Guest
Jon
3 years 3 months ago

I umpire even though it is just Little League, and might I suggest another factor. Umpires changing their calls based on pitcher wildness or accuracy. Many times when officiating I’ll get one kid who obviously pounds the zone but the other guy has the hardest time finding the plate. And as much as I hate to admit it I find a strong urge to widen the zone for the wilder kid and tighten it for the kid with control. Now I know at the major league level things change but I would still think that umpires are faced with similar urges to “balance” the game.
Also umpires are aware that they have some of the biggest impact when the count has 3 balls or 2 strikes, so they are more likely to move away from the spot light and call that strike in a 3-0, or a ball in an 0-2. This may not be that impactful but I would wager that it does have some effect into the data used above.

Member
Krog
3 years 3 months ago

It seems like the opposite happens at the Major League level, with control artists like Tom Glavine getting strike calls on pitches off the plate while pitchers with control problems get squeezed.

Guest
Chris from Bothell
3 years 3 months ago

Commentors above about umpires got me thinking about one of the things that nag at me regarding pitch framing. It really seems like umpires have either been demonstrably worse the last couple years… or able to be under more scrutiny due to instant replay and the internet passing around examples of bad calls… or both.

Pitch framing has to be really hard to get down to an exact measure, with such variance in the competence and consistency of umpires, doesn’t it? I think pitch framing studies do as well as they can to control for the called strike zone vs. actual strike zone, but I’m not confident that the sample size of umpires is wide enough, or the pool of umpires similar enough, to have things even out.

I’m probably phrasing this line of reason clumsily, but: how many home plate umpires are there in a given season? Probably something on the order of 50 – 60, in rotation through the 14-15 games able to go on during any given day of the season. If some of those 50 – 60 umps are worse or less consistent than others, and they’re being a disproportionate part of the total number of “umpire appearances” in a season, doesn’t that throw off the called strikes part of the ‘called strikes v. true strike zone’ part of the equation?

Basically, if the theoretical ideal (that doesn’t just involve robots) is the same human umpire who calls every single game for every single team all season… or one human umpire who is the same all season for each stadium… or one human umpire who is the same all season for each meeting of a team… or one human umpire who is the same for each start by a specific pitcher/catcher tandem… or anything in that progression but a strong measurable amount of “here’s the clearly good umpires, and here’s the clearly bad ones, and the vast majority are in the middle”… well, the further you get from each step of that progression, the noisier the data gets, no?

Guest
marc w
3 years 3 months ago

It’s noisy, sure, but at least Marchi’s work took the specific umpire into account, so the pitches gained/lost take that umpire’s typical zone into account.

As you say, controlling for pitcher, batter handedness, umpire, etc. gets you a baseline with less data, but for the record, someone controlled for each umpire and still got huge framing run values.

Guest
Chris from Bothell
3 years 3 months ago

Oh, ok, thanks. I figured it was going to be noisy and not able to be made – um, quieter? – but good to see that the trends are strong enough to be useful regardless.

Guest
craig richards
3 years 3 months ago

There you go. It would seem to be absolutely impossible to control the variable of pitch location as judged by, what, 50-60 always rotating men during the course of the season. The electronic “tracer technology” is the only way to go to really get the zone “right”. Otherwise, every umpire will bring a slightly different point of view to the calling of balls and strikes. Due to differences in physical size and differences in seeing (visual perception), to say nothing of mental/emotional state (let’s pretend umpires aren’t human, not subject to blurred vison, fights with S.O.s, low blood sugar, or a hangover), this highly subjective experience of slightly differing Men (what, no women?), the strike zone remains a constant ebb and flow of subjective experience.

Guest
brendan
3 years 3 months ago

Hi jeff,

Is it possible to directly count the number of out-of-the-zone strike calls that good framers ‘influence’ and also count the number of in-the-zone ball calls that bad framers ‘influence’? That seems like a good way to measure framing ability, doesn’t it?

I am totally ignorant of the data-mining issues here, but I have read articles that included pitch counts within sub-regions of the strike zone, so it seems like somebody could do this study.

Guest
DRDR
3 years 3 months ago

What’s the formula for Diff/1000? Looks circular to me.

If the expected strikes (E) = Zone%(Z)*total_pitches (T). Then if Diff/1000 (D) = K’s (K)/E, then D = K/(Z*T), or Z = K/(TD). Assuming K/T is roughly similar between catchers, Z = (something random) 1/D. You’ve scaled D to league average, but it looks about what you’d expect, if D is between .5 and 1.5, it will be roughly linear (by eyeball about -1.25)?

Guest
Andrew
3 years 3 months ago

There’s a simple explanation that I’ve seen continually overlooked regarding pitch framing, or any other study with a heavy emphasis on pitch location, zone percentage, etc.: Pitch F/X isn’t a perfect system, and the data starts losing usefulness when measured at such a fine level.

It’s not perfectly damning proof, but other articles and studies have hinted at variability in measurement from one ballpark to the next (e.g. KC very likely has hot velocity readings that would seem to indicate issues with the measurements from which their velocities derive). This study examines 2012 FX fastball classifications and pitch movement from Tropicana field:
http://www.beyondtheboxscore.com/2013/2/13/3968414/pitch-fx-park-effects-case-study-tampa-bay-rays-fastballs-sabermetrics

While the author spends much his focus on pitch movement and type classification, his thoughts are just as relevant to location data. To me, the meatiest part of the study is the discussion of how Pitch FX would likely be set up and calibrated, as well as the factors that are difficult to account for.

Finally, the available pitch FX XML data gives fixed values for pitch movement and location. That data probably can’t be perfectly accurate, however, as the pitch trajectory data for any given pitch doesn’t truly cover release point to crossing the plate. Calculation of those fixed points would require multi-variable calculus, and I don’t believe FX captures all the necessary inputs for perfect algorithms (let alone assuming perfect data capture when there are so many variables).

While FX is remarkable, I think the sabermetric community places a bit too much faith in the unverifiable accuracy of the FX system, particularly when examining pitches at such fine levels. While we like to examine baseball with numbers, we often don’t fully examine the validity and reliability of our data sources. As a result, incomplete or inaccurate data will often lead to conclusions that are themselves incomplete, inaccurate and/or misleading (e.g. defense has become re-emphasized by sabermetricians as a result of the proliferation and evolution of defensive data, while earlier models de-emphasized or ignored it completely).