The proliferation of Stuff fashions has invariably pitted capital-S Stuff towards command when it comes to their respective significance to a pitcher’s success. In case you had to decide on one, is it higher to find nicely, or is it higher to be filthy?
The reply to this query is why baseball is gorgeous and pleasant but in addition often horribly irritating. From Vicente Iglesias and Scott Powers:
Significance, in layman’s phrases, means how related a variable is to explaining an noticed consequence; reliability is how a lot we will count on a variable to repeat or be repeated. In baseball, we frequently speak about these concepts when it comes to “descriptiveness” and “predictiveness.” This dichotomy illustrates the Catch-22 that eternally propels baseball ahead in all its unsure glory: pitch location performs an outsized function in figuring out the outcomes we witness, but we can’t count on to depend on it yr over yr. In the meantime, Stuff is rather more dependable subsequent yr — i.e., modifications to a pitcher’s pitches when it comes to velocity, motion, and launch level are likely to range to a a lot smaller diploma over time — but it surely performs a considerably undersized function in influencing precise outcomes this yr. Brutal stuff. Why will we even hassle?
A part of this may be attributed to the risky nature of the noticed consequence itself. Greater than 700,000 pitches are thrown every year, but solely one-fourth of them terminate a plate look and thus file an official consequence. If a pitcher makes higher pitches within the different three-fourths of his pitches however goofs up those that matter (from an on-field occasion standpoint), he’s out of luck. Moreover, of the one-fourth of pitches that do efficiently finish a plate look, greater than two-thirds of these are batted balls which might be extremely variable of their composition (i.e., exit velocity, launch angle, spray angle) and their worth (e.g., single, double, triple, house run). Usually, the composition of the batted ball is at odds with its worth (i.e., anticipated versus precise wOBA).
That is form of a protracted preamble merely to say: I’m not satisfied pitch location is as unrepeatable as we make it out to be. Intuitively, I feel most of us would consider this to be true. We depend on warmth maps to see how pitchers find their choices; that they present any density, any “warmth,” to start with suggests a capability to repeat. It’s the overly granular nature of pitch location measurements that’s the issue: Statcast measures vertical and horizontal places every to the closest one-hundredth of a foot. To 2 decimal locations! That’s delightfully exact for researchers however dreadfully exact for a pitcher being requested to repeat an assortment of very narrowly outlined places. Even rounded to the closest inch, that is nonetheless an excessive amount of to ask.
So perhaps we have to ask much less. Beneath the present framework, location is unrepeatable. And I’ll settle for that to be true, as a result of, by definition, it’s. However command — which I’ll use at present to explain an array of very exact places that, when aggregated, culminate in a generalized set of targets or zones — shouldn’t be unrepeatable. And whereas we will’t know a pitcher’s supposed goal, I feel we will use a desired goal as a proxy for that. (Simply final week over at Baseball Prospectus, Stephen Brown used fancy modeling to deduce supposed targets for Blake Snell, our enigmatic, team-less Cy Younger winner. Steve concluded the bad-walk-rate-but-good-command argument is partly cognitive dissonance: Snell certainly has below-average command, identical to his double-digit stroll charge suggests, and will afford to seek out the periphery of the zone a bit extra usually. Regardless of Steve’s work, I can’t be inferring supposed targets — we’re sticking with desired targets at present. Sorry, however I’m simply not as sensible as Steve.)
In Statcast knowledge, MLBAM gives a area known as “Delta Run Expectancy,” which is the change in run expectancy after every pitch. (I may do that myself, however I’m glad Statcast did it for me.) I calculated the common Delta Run Expectancy for each pitch location at (x, z) coordinates rounded to the closest inch for all pitches that didn’t produce a batted ball occasion. For batted ball occasions, I calculated the common anticipated wOBA (xwOBA) on contact, additionally provided by MLBAM. All of those calculations are carried out inside Statcast’s pitch kind classifications, for higher or worse, as a result of not all pitch sorts play up (or down) in the identical places. These values are then weighted by the frequency with which a strike, ball, or batted ball occasion happens at each (x, z) pitch location, offering a mean run worth per pitch. I’ll name this “Location Rating.” Right here’s Location Rating by pitch location to the inch, the place crimson is beneath 0 runs (good) and blue is above (regular to unhealthy):
On common, run expectancy is above zero and, thus, blue. If it weren’t, we wouldn’t count on runs to be scored! So it’s a lot rarer to have a competitively commanded pitch that, on common, produces a Location Rating beneath 0. (Please notice that the graphic above flattens each pitch kind’s “scorching zone” into one; it’s purely for demonstrative functions. Every pitch kind’s distinctive “scorching zone” would invariably look a bit completely different.)
I made very deliberate selections with bubble sizing right here. The bigger bubbles symbolize the Statcast-defined “shadow” zone, whereas the tiny dots symbolize all the pieces else. Throughout the shadow zone, I delineated the “interior” shadow (bigger bubbles contained in the zone) from the “outer” shadow (smaller bubbles exterior the zone). You possibly can see how nicely Location Rating corresponds with commanding inside the interior shadow. There are nonetheless pockets of interior shadow that don’t correspond nicely, particularly on the center of the east, west, and south edges of the zone. Eradicating these places, it doesn’t go away a number of daylight for commanding pitches actually nicely. However, by and enormous, it typically reconciles properly with the interior shadow.
From right here, I calculated a easy binary: Did the pitcher command his pitch in a scorching zone (crimson, beneath zero) or a chilly zone (blue, above zero)? A easy common of those zeroes and ones yields a proportion of pitches {that a} pitcher commanded competitively or (comparatively) uncompetitively. (Ideally, this gels with PitcherList’s Mistake Price metric — actually, I could also be re-creating the wheel with this entire endeavor — though I’m loath to confess I didn’t overcome my laziness to really examine.)
For reference, the league-average charge of aggressive pitches is 35.5%, which is actually increased than I anticipated. Right here’s a desk of the highest 15 pitchers who threw at the least 1,500 pitches in 2023. Most of the names gained’t shock you within the slightest, and a few undoubtedly will. The second column is the pitcher’s charge of aggressive pitches on the final pitch of a plate look, that specific and finicky subset of outcomes upon which a pitcher’s efficiency is definitely evaluated on its face:
% Aggressive Pitches Above Common
I gained’t rattle off the names of the non-surprises. How about Hunter Brown, Yusei Kikuchi, and Nick Pivetta? Kutter Crawford, perchance? It’s actually attention-grabbing (to me, at the least) to see who excelled when it “mattered” (Joe Ryan, Justin Verlander, Gerrit Cole, Kevin Gausman) and who didn’t (Julio Urías, who notably struggled). We’ll get to the significance and reliability (a-ha!) of those measurements in relation to one another shortly.
However first, let’s get to the laggards. Right here’s 2023’s backside 16 (one bonus title, as a result of it’s topical):
% Aggressive Pitches Under Common
I feel we will shortly establish an anecdotal argument that Location Rating unduly punishes excessive sinker/changeup guys. They depend on pummeling the underside of the zone, usually beneath the underside fringe of the zone. That’s superb. I imply, it’s not superb, however I see it, and we’ll handle it by some means, someway, sometime. However for now, it’s what it’s. It’s necessary to recollect, too, that this complete train is agnostic of Stuff — and guys like Framber Valdez, Logan Webb, and (even at his worst) Corbin Burnes have spectacular Stuff.
(My idea for why that is, by the best way, is Statcast’s xwOBA could be very a lot batter-centric, such that it emphasizes exit velocity rather more than launch angle. Nevertheless, pitchers have rather more management over launch angles than exit velocities. The answer, then, can be to concoct another pitcher-centric xwOBA and insert it into this evaluation. I feel it could resolve this situation with heavy sinker/changeup guys particularly, and inevitably different unseen contact administration points, too.)
I’m not right here to diagnose every pitcher’s Location Rating plus-or-minus common; I’m simply right here to check reliability. So! The large query! Is command, by advantage of Location Rating, dependable — that’s, is it repeatable yr over yr? You wager your candy cheeks it’s, pal. Listed below are the chances of aggressive pitches in 2022 (x-axis) and 2023 (y-axis) for all pitchers who threw at the least 500 pitches in every season (n = 397):
An r-squared of 0.43! That’s very stable. Moreover, the within-year relationship for 2023 (n = 276 pitchers, min. 500 whole pitches) between intermediate (non-PA-ending) and remaining (PA-ending) aggressive pitches is equally robust (r-squared = 0.47)…
… which is robust sufficient for me to imagine, maybe mistakenly, that massive variations between a pitcher’s aggressive pitch charge to finish plate appearances should regress towards his common aggressive pitch charge for all pitches. It’s attainable some pitchers don’t do that, that some — like Verlander and Cole among the many recreation’s perennial elite, or, on the different finish, Cal Quantrill — persistently command their pitches higher or worse in higher-leverage counts. I completely consider that. However, by and enormous, I might count on most pitchers don’t, or can’t.
And that’s all I’m right here to say. Location shouldn’t be repeatable, however I feel command (as I’ve outlined it right here) is.
It’s necessary to acknowledge that variations in command-related abilities quantity to only some proportion factors above or beneath common amongst all pitches thrown. It’s actually not loads — though, over the course of a single recreation, Ryan making seven further locationally aggressive pitches than the common pitcher shouldn’t be one thing to sneeze at. It’s simply that, of these aggressive pitches, what occurs to them? Are they intermediate or remaining pitches? Once they finish a PA, do they permit a batted ball? What does that batted ball appear to be? Was Ryan “fortunate”? “Unfortunate”? There’s simply a number of variables — and, thus, a number of variance — to confront.
That’s neither right here nor there for now. I simply wished to make a case for command as not solely a ability {that a} pitcher owns but in addition a ability that may be relied upon yr over yr. To be clear, I’m not saying anybody doubted this. I don’t need to make it look like there was some consensus about command being, like, not actual. I do assume, although, that its significance received misplaced in translation with the popularization of Stuff fashions that emphasize pitch form over pitch location — and that accompanying pitch location fashions, with the hyper-specificity of pitch location coordinates, might lose sight of the forest for the bushes. Permitting pitchers the grace of a broadly outlined desired (or, if obtainable, supposed) goal helps resolve that.
So: In case you may select one, is it higher to find nicely or just be filthy? The reality is, you don’t have to decide on. You possibly can most likely have each — reliably.
