- FanGraphs Baseball - http://www.fangraphs.com/blogs -

Pitch Receiving and Pitch Types

I think I’m the guy at FanGraphs who’s most interested in the field of pitch-framing research, so, hey there, here comes another post about pitch-framing. The idea of the importance of receiving a pitch correctly has been around forever, but only more recently have people begun to feel like they can measure who’s good and who’s bad at it. You know the good ones are represented by Jose Molina, and you know the bad ones are represented by Ryan Doumit. There are calculations that use PITCHf/x data to figure out strikes above or below average, for catchers.

What isn’t quite so clear, yet, is how we should interpret those numbers. Catchers aren’t all catching the same pitchers in the same ballparks in the same counts with the same umpires and the same opposing batters. It’s complicated, because of course it’s complicated, and I’m interested in the idea of some guys being harder to catch than others. Some guys will be more and less easy to frame. At its core, this is because some pitches should be more and less easy to frame. Catching a 95 mile-per-hour fastball is going to be different from catching a 70 mile-per-hour lollipop curve.

Following is some initial research into what we can think of as strike zones by pitch. PITCHf/x already provides pitch classification, and while the classification is imperfect, it’s gotten better and it’s close enough over big samples. We can tally up how many of each pitch type there were, and we can tally up how many of those pitches went for strikes. We can also tally up how many of those pitches were in the strike zone, and how many of those pitches resulted in one-of-zone swings. From here, we can end up with a number of expected strikes, which we can then compare to the number of actual strikes.

That’s what you’re going to find in the table below. With help from my friend and co-writer Matthew Carruth, I identified all pitch types thrown at least 1,000 times in 2012. The column “Diff/1000” refers to actual strikes minus expected strikes, per 1,000 called pitches. A positive number means there were more strikes than expected, while a negative number means there were fewer strikes than expected. The overall average here is pretty much exactly 0. More strikes than expected implies easier pitch receiving; fewer strikes than expected implies just the opposite. The data:

Pitch Count Diff/1000
Sinker 64205 6
Two-Seam 91011 5
Four-Seam 238423 5
Cutter 39013 4
Splitter 9283 0
Changeup 71125 0
Slider 109157 -6
Fastball 1514 -10
Curveball 69790 -13
Knuckleball 2799 -26
Knuckle Curve 1114 -49

A few things to explain, first. The knuckleballs are all R.A. Dickey knuckleballs, so it’s not like we’re drawing from a big, averaged-out sample. That’s biased by the particular pitcher and the particular catchers. The knuckle curves are mostly A.J. Burnett knuckle curves, so again, we’re biased by the pitcher and the catchers. That generic “fastball” listing isn’t grouped with the other fastballs, and we’re probably seeing small-sample noise, there. A sample of 1,514 pitches is a small sample, for these purposes.

The results are probably about what you’d expect. Fastballs are the easiest pitches to receive, getting the more favorable strike zones. There’s little difference between the sinkers, the two-seamers, the four-seamers, and the cutters. As pitches move more, we see less favorable strike zones. Of course it’s challenging to catch a knuckleball — nobody knows what the pitch is going to do. Of course it’s challenging to catch a big curveball, and maybe especially an A.J. Burnett curveball. Catching those pitches might require more movement on the catcher’s part, and of course the break can be deceiving. Pitches with more movement will be more difficult to locate, too, and there’s presumably a strike-zone benefit to pitching around your spots instead of unpredictably all over the zone.

What isn’t controlled for in here are the counts. Breaking balls are thrown in a lot of pitcher-friendly counts, and pitcher-friendly counts have tighter strike zones. The opposite is true for fastballs. That’s going to be a factor, although it’s unlikely it negates all the differences we observe.

A guy who throws a lot of sinkers is Brad Ziegler. Over the last few years, Ziegler has had one of the most favorable strike zones in baseball. There could be a relationship there, and here’s a .gif example of Ziegler getting a borderline call:

Meanwhile, here’s a .gif example of A.J. Burnett not getting a borderline call on a knuckle curve:

On their own, the two .gifs don’t prove anything, but people always like looking at images instead of text. The surprises in here are few and far between. It seems like it’s the easiest for catchers to receive fastballs. It seems like it’s the hardest for catchers to receive breaking balls. There’s more work to be done, here, more steps we can take to get better ideas of what’s going on. How about fastballs that move in different ways? That’ll be a post for another day, probably. Same with fast curves and slow curves, probably. Lots of potential divisions, and who doesn’t like dividing?

There’s a lot that goes into determining what kind of strike zone a pitcher will have. Each and every one of those factors is absolutely fascinating.