Why We Still Don’t Have a Great Command Metric by Eno Sarris April 25, 2017 To start, we might as well revisit the difference between command and control, or at least the accepted version of that difference: control is the ability to throw the ball into the strike zone, while command is the ability to throw the ball to a particular location. While we can easily measure the first by looking at strike-zone percentage, it’s also immediately apparent that the second skill is more interesting. A pitcher often wants to throw the ball outside of the zone, after all. We’ve tried to put a number on command many different ways. I’m not sure we’ve succeeded, despite significant and interesting advances. You could consider strikeout minus walk rate (K-BB%) an attempt, but it also captures way too much “stuff” to be a reliable command metric — a dominant pitch, thrown into the strike zone with no command, could still earn a lot of strikeouts and limit walks. COMMANDf/x represented a valiant attempt towards solving this problem by tracking how far the catcher’s glove moved from the original target to the actual location at which it acquired the ball. But there were problems with that method of analysis. For one, the stat was never made public. Even if it were, however, catchers don’t all show the target the same way. Chris Iannetta, for example, told me once that his relaxation moment, between showing a target and then trying to frame the ball, was something he had to monitor to become a better framer. Watch him receive this low pitch: does it seem like we could reliably affix the word “target” to one of these moments, and then judge the pitch by how far the glove traveled after that moment? How about all those times when the catcher is basically just indicating inside vs. outside, and it’s up to the pitcher to determine degree? What happens when the catcher pats the ground to tell him to throw it low, or exaggerates his high target? There are more than a few questions about an approach affixed to a piece of equipment, sometimes haphazardly used. The results of COMMANDf/x were also interesting from another viewpoint. They suggested that the average pitcher missed his spots by 13 inches, and that even the best pitcher (Dallas Keuchel, in that particular sample), throwing a fastball on a 3-0 count, missed the glove by almost 10 inches. Most catchers and pitchers who’ve heard this factoid from me have laughed and then expressed their disagreement in some form or another. You could take away from those last numbers that command is overrated and we should focus on stuff. That might be a good takeaway, considering how difficult it is to command a pitch. One degree of difference in your release angle on the mound means a foot of difference at the plate. Even if we do our best, our bodies and our mechanics are complicated enough to make multiple degrees of difference upon release an inescapable reality. Or maybe we need to get better at measuring command. Baseball Prospectus has a great stat called Called Strikes Above Average that attempts to strip out all sorts of context to see the what effect the pitcher has on the probability of a called strike. That’s awesome! It even strips out the effect of the catcher’s framing, which is important to all of this. One thing it doesn’t do right now, though? Account for the swinging strike. Some pitchers get called strikes with command, some use their command to coax a swing out of the better. One method that I’ve found compelling in the past is Bill Petti’s Edge statistics. He described a location that was the width of a baseball inside the strike zone as the edges and then tracked which pitchers were able to hit that location. The latest rollout of new Statcast stats and search functions has updated those Edge stats to create a new zone that also includes a baseball’s width on the outside of the zone. Enlarging the target area to just outside as well as inside the zone makes sense. You often want to put that ball just outside the batter’s reach, especially on a swinging-strike count. So, let’s look at the list of pitchers who can best put the ball within a ball’s width of the strike zone, and let’s limit it to counts where the pitcher is ahead. Presumably, in those cases, the pitcher isn’t trying to throw to the middle of the zone to get a strike. Best Edge-Throwers in 2017 Player Results Total Pitches % of Pitches Michael Pineda 79 333 23.7 Clay Buchholz 33 144 22.9 Kyle Gibson 69 303 22.8 Jason Vargas 64 283 22.6 Matt Shoemaker 84 376 22.3 Rick Porcello 91 414 22.0 Chris Sale 93 425 21.9 Tyler Skaggs 80 371 21.6 Joe Musgrove 79 367 21.5 Aaron Nola 59 275 21.5 Brandon McCarthy 75 354 21.2 Dan Straily 71 336 21.1 Kyle Hendricks 58 276 21.0 Rich Hill 27 129 20.9 Tommy Milone 61 292 20.9 Jameson Taillon 78 375 20.8 Matt Andriese 78 377 20.7 Chase Anderson 73 354 20.6 Clayton Kershaw 78 381 20.5 Marco Estrada 82 403 20.4 Francisco Liriano 44 217 20.3 Counts: 0-1, 0-2, 1-1, 1-2, 2-2Zones: within a ball’s width of the edge of the strike zone in all four quadrants It’s a weird list. Anyone who watched a Michael Pineda start last year — including our own Nick Stellini — would have sworn that command was his problem. Pineda, of course, produced excellent strikeout and walk rates. He also allowed way more runs than his fielding-independent numbers would have suggested he should. Why? Because he allowed loud contact when he wasn’t missing bats. The sort of loud contact, you’d think, that results from missing a spot. Perhaps, then, it’s because this edge-throwing methodology doesn’t account for the fact that Pineda’s misses are middle-middle? Pineda was 167th out of 238 in hitting the very middle of the zone in these counts, and is 67th out of 73 so far this year. I don’t think that’s it. He threw his slider middle-middle 37th-most out of 117 qualifiers last year. Is that it? Slider command? Is it possible that Pineda can miss bats, limit walks, hit the edges better than almost anyone… and yet lack great command? And then there’s the relationship of stuff and command in the other direction. If you don’t have great stuff, you have to live closer to the zone and try to get called strikes. Do Kyle Gibson, Jason Vargas, and Tommy Milone have great command, or are they nibblers who have to take little picks and pecks at the zone and make sure they get no closer? Would it change your mind if you knew that nobody has a higher edge-of-zone minus heart-of-zone percentage this year than Gibson? Maybe, in this case, nibbling is command, and it’s a lack of raw stuff that’s hurting us from understanding it as such. Our sniff test is all we have in these cases, and while it’s nice to see Clayton Kershaw on here — he’s the first name out of any pitcher’s mouth these days during a discussion on command — there’s also Francisco Liriano. Nobody’s paragon of command. I don’t know how many times I’ve stifled a cuss word during a Rich Hill start, when he can’t find the zone with his fastball — even while he’s hitting his spots with his curveball. So that brings up fastball vs. secondary command, which can greatly vary from pitcher to pitcher. Most pitchers have better fastball command than breaking-ball command, it’s just Hill alone on that mountain. It’s tempting to zero in on fastball command, but even when you re-do that list from above for only fastballs, you’ll get your head scratchers (Vince Velasquez and Tyler Glasnow?). Maybe they were aiming for low and inside and it ended up low and outside. The core problem may be unsolvable. In order to best judge command, you’d have to know exactly what the pitcher intended to do with the ball. Without being able to ask each pitcher, we’re left guessing at their intent, and that’s what causes our issues. Is intent where the catcher puts his mitt? Can we use the count to guide our attempt to understand intent? Is command the same for different pitches? It seems the best approach would account for as many factors as possible, and it might even need to be run by humans. It might not be measured in inches missed or balls put in a certain zone. Maybe it’s best treated as a binary question! Did the pitcher achieve his intent or not? Perhaps a human could score a game with a few designations — things like hung (secondary pitches too high in the zone on swinging counts), yanked (dirt or low), way off (completely different zone than intended), just missed, missed, and nailed — that he can put on each pitch. A human can watch and say something like “Well, it was clear he wanted that outside and he got it outside but not far enough, or too far” versus “Yeah that was low like he wanted, but low and inside instead of low and outside.” Perhaps we’d differentiate between “missed high/low” and “missed inside/outside.” Perhaps we’d be able to talk about a pitcher’s command to different sides of the plate, and differentiate between guys who hang secondaries and those that miss vertically or horizontally. Without that human to judge the pitch, it’s really hard, if not impossible, to figure out exactly what that pitcher meant to do. And that’s a prerequisite for judging command.