What Happens When a Pitcher Gains or Loses a Framer?
To be honest, this might be a post I’ve already written before. It feels like that sort of thing. But I’ve got pitch-framing on my mind, again, and even if I have done this before, at least now we’ve got more data. Why not revisit something whenever you have more data?
When we talk about framing, so often the focus is on the catchers. This is appropriate because it’s the catchers who are doing the framing — sometimes with a little assistance from other sources. But it’s the pitchers who are actually affected, because it’s the pitchers throwing the balls, and the catchers are involved only after the rest of the play has been carried out. It’s not entirely clear how pitchers are impacted by their receivers. We have estimates, we have some pretty good ideas, but the reality is complex. In this post, I try to examine something simple: What happens when a pitcher goes from better receiving to worse receiving? What happens when a pitcher goes from worse receiving to better receiving? As is literally always the case when I run an investigation, there are better and more thorough ways to do this, but I’m woefully limited by my own lack of ability, and if I don’t have the quick-and-easy stabs, I don’t have anything.
So, once more, I’m going to the well of the framing statistic I made up some years back. The quick, usual refresher: Based on data we have at FanGraphs, you can calculate a pitcher’s expected strike total, and then you can compare that to the pitcher’s actual strike total. This gives you an idea of how well the pitcher was or was not received. It isn’t perfect — few things this simple can be said to be perfect — but it seems to hold up pretty well, and the results tend to match what you’d expect given what we understand about different catchers.
When I run these calculations, I use the PITCHf/x plate-discipline data we have going back to 2008. While we also have other plate-discipline data going back to 2002, I find this data to be unreliable, so I don’t use it. There’s nothing wrong with, say, the calculated contact rates from 2004, but I don’t buy the zone and out-of-zone numbers. So I leave those be. It’s a shame, but I haven’t yet figured out what might be possible with that old and flawed information.
I selected starting pitchers who threw at least 100 innings in consecutive years between 2008 and 2014. For each year, I calculated the expected strikes compared to the actual strikes, and after putting these over a common denominator, I was able to see which pitchers had the biggest season-to-season increases and decreases in extra-strike rate. I interpreted big increases and decreases as representing improvements or declines behind the plate. This isn’t always true, but, it’s what I’m going with. I sorted the pitchers in descending order of extra-strike improvement, and then I placed them into groups.
Group 1 pitchers saw an extra-strike-rate increase of at least 1 standard deviation.
Group 2 pitchers saw an extra-strike-rate change in between +/- 1 standard deviation.
Group 3 pitchers saw an extra-strike-rate decrease of at least 1 standard deviation.
With the pitchers grouped, all that was left was comparing season-to-season statistics. In the table, you can see average changes between Year 1 and Year 2.
| Groups | WAR/200 | RA9/200 | ERA- | FIP- | xFIP- |
|---|---|---|---|---|---|
| Group 1, Y1 | 3.1 | 3.0 | 95 | 94 | 93 |
| Group 1, Y2 | 3.2 | 3.1 | 93 | 93 | 94 |
| Change | 3.1% | 4.3% | -2.4% | -0.9% | 0.6% |
| – | – | – | – | – | – |
| Group 2, Y1 | 2.9 | 3.0 | 95 | 96 | 97 |
| Group 2, Y2 | 2.6 | 2.5 | 100 | 99 | 99 |
| Change | -11% | -16% | 5.2% | 3.4% | 1.9% |
| – | – | – | – | – | – |
| Group 3, Y1 | 2.8 | 2.6 | 99 | 96 | 95 |
| Group 3, Y2 | 2.4 | 2.3 | 101 | 101 | 100 |
| Change | -14% | -10% | 2.7% | 4.3% | 5.2% |
Group 2 might be the “control” group — the group that didn’t meaningfully change in terms of extra strikes. You can see that, in Year 2, the average pitcher in Group 2 got worse. This is what you’d expect; there’s a little age-related decline, and there’s a little survivor bias. In Year 1, the pitchers, as a whole, were a little better than average. In Year 2, they were closer to exactly average. You see steps back across the board.
Compared to Group 2, Group 3 is interesting because it actually saw less ERA decline, despite getting less help from the catchers. However, there’s more FIP decline, and there’s more xFIP decline. The changes are small, relatively speaking. It has to be pointed out that the groups aren’t composed of identically-talented pitchers; reality just doesn’t work like that. So that’s one potential source of error, as is the foundational calculation being used to show framing help. Also, the sample sizes. The pitchers in Group 3, as expected, performed worse in Year 2. They looked like approximately average starting pitchers, on balance. Compared to the control group, it’s hard to see a change of more than a few runs.
That leaves us with Group 1. Look at the numbers for Year 1: you’d think that, in Year 2, the pitchers would perform worse. That’s how these things go. Instead, the pitchers got ever so slightly better, everywhere but xFIP, which remained effectively the same. This would presumably be the framing effect. Now, it’s also impossible to isolate this from any potential improvements in pitcher command, but while that could be a factor, I’m content to keep things simple. In other groups, the pitchers performed worse. In the group that saw the biggest increases in extra-strike rate, decline was staved off. It isn’t easy for good pitchers to not get worse the following season. These pitchers pulled it off.
But then, examine the differences at the extremes. Based on the framing data and the rule-of-thumb run value of 0.14 for an extra strike, the difference between Group 1 and Group 3 should be somewhere in the vicinity of 15 runs per 200 innings. What we see instead is something like, what, half a win? It’s big enough for me to believe it, but it’s not as big as you might expect, leading me to think there could be other, compensatory factors. We’ve touched on this before from time to time. Pitchers with worse receivers might spend more time in the strike zone, balancing out some balls on the edge. Pitchers with better receivers might spend more time out of the strike zone, generating higher-quality strikes but a greater probability of balls. It seems like positive and negative values are being reduced. That is, the framing effects are there, but the pitchers pitch differently based on the abilities of the guy behind the plate. I might be getting in a little over my head. I’m just following my own numbers.
If you’re curious, the biggest season-to-season increase in extra-strike rate: Hisashi Iwakuma, from 2013 to 2014. His ERA got much worse, but his FIP remained stable, and his xFIP took a step forward. Iwakuma benefited from Mike Zunino’s strike zone.
And the biggest season-to-season decrease in extra-strike rate: Derek Lowe, from 2011 to 2012. The second year, he was dreadful. It’s not much fun for a pitcher to go from Brian McCann and David Ross to Lou Marson and Carlos Santana.
As a closing thought, it’s interesting to think about some guys who might see the biggest changes in receiving help in the year ahead. For pitchers like Yovani Gallardo, Andrew Cashner and Tyson Ross, they might need to try to compensate for worse framers. At the other end, there’s promising news for Marcus Stroman, J.A. Happ, Nathan Eovaldi and Justin Masterson. The framing isn’t everything. The framing isn’t close to everything. But it’s hard to see the downside of a more favorable strike zone. That is, unless you’re hitting.
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.
Drew Hutchison might benefit as much as anybody by the difficult-to-measure pitch-framing effect. He is already a strikeout pitcher, and he lives high in the zone (where umpires seem to be calling fewer strikes lately). If he gets a slight benefit high in the zone from Russ Martin’s framing, he could be a high-end starter. He is not too far away from that level.
Unfortunately for Mr. Hutchison, I believe there was an article showing framing benefiting low pitches far more than high ones.
Yes, but parse that in the latin declension and my point is still moot.
Martin might not help Hutchison’s 4-seamer, but he should really help Hutchison’s newish slider (http://www.fangraphs.com/blogs/how-good-of-a-weapon-did-drew-hutchison-find/).
Ancedotally, I remember hearing that specifically about Martin as well.
what’s interesting about the jays is that JPA ranked significantly better in pitch framing than Navarro, yet they all did much better throwing to Navarro.
I’m still a pitch framing skeptic.
so many variables….what if iwakuma made some tweaks to his delivery that made the difference? what if there was a nagging injury bothering lowe…
Exactly what I was thinking. Generous umpires vs. hitter’s umpires, what parks they pitch in, what league, etc.
Has anyone ever asked the umpires about this framing concept? I would think that they may have an opinion on the supposed effect?
“Pitch Framing” is the art of fooling umpires, and an indicator that computerized strike zones are inevitable. So I bet they probably don’t like that it exists.
So then it makes more sense to talk about banning pitch framing than the defensive shift.
No; because pitch framing contributes to less number of pitches thrown and actually speeding up the game.Good framing means I have to swing at that pitch I might otherwise not.
While there is a base number for innings pitched, could there by variances in these numbers based on the actual number of pitches thrown?
Sure, in the sense that the variance of expected strike rate will be higher for pitchers who threw 100 innings vs pitchers who threw 200 innings. That said, there are a lot of pitches in 100 innings (at least 1200 I would imagine, probably more), so I don’t imagine this will have a huge impact.
Jeff, could you provide the average number of innings pitched in years 1 and 2 for each of the 3 groups?
I’ve got to imagine that Jake Arrieta will like his job a tad bit more this year. Castillo and Buck become Montero and Ross
Buck?
I’m trying to work through the math here.
If the difference between group 1 and group 3 is about 5 runs per 200 innings. The top group (> +1 SD) should have an average z-score of about +1.4 and the bottom group (< -1 SD) an average of about -1.4. So, a 2.8 standard deviation difference in extra strikes. 5/2.8 = an SD of 1.8 runs per 200 extra strikes which would mean that someone 2 SD above average would be +3.6 runs per 200 or +20 runs per 1100 innings. Isn't that what StatCorner is showing with the top catchers about 20 runs above average?
A difference of 15 runs/200 innings between group 1 and group 3 would imply that a +2 SD catcher is +60 runs per 1100 innings.
I accidentally favored my reply but I guess everyone needs to give themselves a pat on the back once in a while.
That said, I meant to write “1.8 runs per 200 inning in extra strikes” above.
I should add that I think the WOWY with pitchers method is a great way to see if these pitch framing numbers hold up in practice. I don’t really understand the extent of the skepticism of catcher framing numbers though. If we know that catchers get extra strikes and that it’s a repeatable skill, the theory that this gets cancelled out or muted by changing approaches strikes me as an improbable one that relies on pitchers having more control of the locations of their pitches than they could realistically have.
The Group 1 pitchers experienced better receiving in Year 2 by a little over 50 extra strikes per 200 innings. The Group 3 pitchers experienced worse receiving in Year 2 by a little over 50 extra strikes per 200 innings. Using the .14 run value, there you’d expect some separation somewhere of about 15 runs. Mucking this up to some extent is that the population pools aren’t identical, so we’re comparing pitchers against different pitchers. Also potentially mucking this up to some extent is my own brain, and the limitations of it.
Thanks. I was misunderstanding this.
It probably wasn’t very clear. And I was just re-reading some old MGL and Tango analysis, and their WOWY studies by and large supported the expected framing-effect ranges. So I don’t really know what’s happening. I defer to their more rigorous studies, but something’s still gnawing at me. I feel like the huge effect should be easier to observe.
So you’re saying changing the number of called strikes a pitcher gets changes their strikeout and walk rate? Truly groundbreaking stuff here.
I think you might want to reread this.
Framing is interesting, but an element to be considered is the ump. Brooks had umpire information showing zones they would call so it can be assumed that if I could discover that different umps consistently called a wide or low zone that a catcher would know this and call and frame accordingly.
Second thought on framing is that there are two ways to set a pitch target, the first put the mitt where you want the pitch to end up, the second where you want the pitcher to aim for before the break. Pitchers with a lot of motion don’t always know where the pitch will finish so throw a spot like with a fastball then let the slide or change end up where it will end up.
A third thing might be that umps get to know catchers and they talk quite a bit during a game, maybe even the catcher signals the ump what and where the pitch is. Catcher sets a target mid-thigh and in and lets the ump know its a slider away, so ump doesn’t get fooled.
Did you happen to look at a sample of pitchers from groups 1 and 3 (beyond Iwakuma and Lowe), and see if most of the pitchers in said groups actually had different receivers on average?
I imagine this will significantly hurt your sample size, but what happens to your analysis if you only include pitchers who switched teams between years 1 and 2?
That’s what I was was thinking. This is just an exercise unless your looking at receiver changes and comparing directly