Austin Brice and the Value of Release Point Repetition

Austin Brice is a legitimate prospect. The Marlins spent $205,000 to sign him out of high school in 2010, and he was ranked as the sixteenth-best farmhand in the Miami organization by Baseball America coming into the 2013 season, an area of prospect lists he will likely to continue to reside in this offseason. He’s just 21, has two pitches that flash plus, and has a prototypical pitcher’s body and smooth, easy, delivery.

He also has 190 career walks in 279 2/3 professional innings, including 82 in 113 frames in 2013. That’s a career 14.88% walk rate and a 15.16% mark in 2013, a number that was actually a step back from 2012 (14.08%) even though he was repeating the Low-A level (his ERA also shot up from 4.35 to 5.73, and his K-rate fell from 25.26% to 20.52%. Certainly, this past season did not bring the young righthander much good news.

Plenty of pitching prospects pair tantalizing stuff with frustrating inabilities to throw strikes, but Brice (whom I saw five different times in 2013, a virtue of living 45 minutes from NewBridge Bank Park) is an especially frustrating case because, as I said above, his delivery is one of his strengths. In this piece, I’m going to examine the root of his control problems and tie it to some more general and important lessons about the process behind throwing strikes.

But let’s begin at the beginning. Austin Brice is a prospect. Three reasons why he is a prospect are very clear in one simple .gif:


You can see here that:

1.) Austin Brice looks like what a pitcher is supposed to look like. He’s listed at 6’4″ and 205 pounds, and he has a well-proportioned build that looks like it could hold maybe a bit more muscle as he develops.

2.) He has a really nice, easy delivery. He stays tall and gets good plane even from a 3/4 arm slot, the motion is very smooth and low-effort, the arm action is clean, and he’s very compact–there’s not much wasted motion at all.

3.) The above pitch went 95 mph.

That’s a great collection of attributes to have at 21–a durable, athletic build, a classic, sound motion, and easy, effortless velocity. The heater might be a bit on the straight side, but he started incorporating a two-seamer late in the year to help him out in that regard. He also can unleash this monster:


I saw something on the order of 80-90 minor league baseball games this year, and I’m not sure I ever saw a pitch that had sillier action than this 76-mph curveball. No, it’s not always that good, and the fastball’s not always 95, but the fastball averages a solid 92+ and the curve works in the upper 70s with bite. Brice also has a changeup that’s best described as “playable”–it lags far behind the fastball and curve, but he’s not afraid to use it, and it gives hitters a third distinct speed to worry about.

In the above .gif, you can see Brice from the stretch. He’s very similar to how he looks in the windup–low-effort, little wasted motion, with good leverage to the plate. I wish more minor league pitchers would use deliveries this clean and repeatable.

If you never had heard of Austin Brice, and all you saw was these two .gifs, you probably never would guess that he walked over 15% of the batters he faced this year, but indeed, he led the South Atlantic League in walk rate by over a full percent. There is a profound inconsistency between Ideal Austin Brice, the pitcher we see in these .gifs, and Statistical Austin Brice, he of the massive walk problems, and it makes him an extremely frustrating prospect to wrap one’s head around. When we see a bad number on the stat sheet, we expect to be able to watch the player and pick out some major flaw that leads to that bad number. With Brice, it’s a little more subtle.


Projecting command and control abilities for minor league pitchers is a tricky business. Most analyses of pitching prospects tend to focus on stuff. If a guy has great stuff, we figure, the subtleties of pitching will sort themselves out later–pitchers have years to figure out how to find the strike zone, whereas a pitcher can’t really “learn” his way to plus velocity or movement. Of course, finding the zone–and putting one’s stuff to optimal use–is an integral part of pitching–nobody denies that. But, other than citing things like walk rate and K/BB ratio, there isn’t that much delving into the specifics of a pitching prospect’s command/control projection that goes on in analytical circles.

The numbers are a nice start, but particularly in the low minors, they really aren’t of too much use. Chris Archer had Brice-esque walk rates for almost all of his minor league career, and yet, in his rookie season in the big leagues, suddenly he was walking batters at a lower-than-average rate. On the flip side, this year’s MLB walk rate leader was Pittsburgh’s Jeff Locke, who always had walk rates between 3.4% and 6.2% in his first five years of minor league ball, including a stellar 4.9% in his 2010 introduction to Double-A. You don’t have to look far to find plenty of other examples of this phenomenon.

So, how do we get a more nuanced picture of how a minor league pitcher might fare in the control department? Personally, I project it largely based on a pitcher’s delivery. The sounder and easier the motion is, the easier it’s going to be for the pitcher to find his target with consistency. That makes sense–intuitively, when we think of masterful control pitchers, guys like Roy Halladay, Cliff Lee, Bartolo Colon, or Adam Wainwright come to mind. All of those players have easy motions that they seem to just lock into a groove with and repeat effortlessly.

There’s also the matter of “control” versus “command.” The way most people differentiate the two terms is “Control is throwing strikes; command is throwing quality strikes.”

But that’s neither a useful nor a sensible distinction. In this definition, “control” is differentiated from “command,” presumably, by the amount of “non-quality” strikes a pitcher throws. And yet, “control” is still treated as a vaguely positive term, as if throwing large amounts of non-quality strikes is a) a good thing and b) represents successful execution of the pitcher’s plan. Conversely, “command” involves throwing lots of quality strikes, the most common of which are presumably pitches on the edges of the strike zone. But pitchers universally believed to have good “command” do not always do this.

Consider the final three years of Livan Hernandez‘s career. Livan’s fastball averaged right around 84 mph, he had a fairly nondescript slider and change, and a big bloop curve. His delivery had no deception to speak of. And yet, he survived, having two pretty solid years before finally melting down in 2012, on the back of his command. How else do you explain how he tamed MLB bats with, at best, 30-grade stuff? Hernandez had good walk rates, too–the statistical evidence was there.

Except Hernandez wasn’t throwing “quality strikes”–he wasn’t throwing strikes at all. In each of those seasons, he threw well under 40% of his pitches in the strike zone, and was 7.5% or more below the league average in zone percentage each year. What he did instead was throw a lot of pitches just a few inches outside of the zone–major league hitters can still hit 84-mph heat or a 66-mph curve if it’s three inches off the plate, but that’s a tough enough location that the contact wasn’t likely to be particularly high-quality. All Livan had to do was induce a swing before a batter took four borderline pitches. Of course, the reality of his approach was more complex than that, but the point is that Livan had good “command” in a subjective sense, but not a statistical sense.

With this in mind, I think the important things to look at with a pitcher’s control attributes are not “strikes” and “quality strikes,” but rather “zone tendency” and “hitting spots.”

The nice thing about framing it this way is that it takes the discussion from what the results were (Was the pitch a strike? Was it in a good location?) to a focus on the process. Where pitches end up is valuable information, but when it comes to analyzing the pitcher, it’s even more important to know why pitches end up where they end up. Zone tendency tells us something about what the pitcher’s plan is, whereas hitting his spots tells us how well he can execute that plan. Livan Hernandez had a low zone tendency but survived because he did a great job hitting his spots. Cliff Lee has the same excellent ability to hit his spots, but he relentlessly pounds the zone. Edinson Volquez works in the strike zone a slightly above-average amount, but doesn’t hit his spots very well.

If that’s how we should think about control/command (I propose that the present term “control” be scrapped and replaced with “zone tendency” whereas “control” and “command” should be synonymous and refer to a pitcher’s ability to hit his spots) in a present-day sense, then how can we think about projecting it? A pitcher’s plan certainly can evolve over time, especially as he progresses up the minor league chain and faces different calibers of hitters that force him to adjust, but there are still some tendencies that can be observed that have some predictive value. In particular, what a pitcher can do location-wise is probably more noteworthy than what he can’t. For example, in my last post, I talked about how Tony Bucciferro showed the ability to work all three of his pitches inside to lefthanders, something many righthanded pitchers can’t do. It seems less likely that Bucciferro will lose his ability to work inside to southpaws than it is for another pitcher who doesn’t presently do that to learn how to.

Then there’s the matter of executing one’s plan, and here’s where the mechanics come in. There is, of course, a ton of virtual ink spilled about pitching mechanics, and there is not a total consensus on what works and what doesn’t. Some pitchers have deliveries that make everyone grimace and still find a way to hit their spots, whereas others…well, others find themselves in Austin Brice‘s conundrum.


Austin Brice‘s delivery is both technically sound and aesthetically pleasing, but neither technical soundness nor aesthetics get a baseball to its desired location. In order to execute a pitch, the most important thing is executing the delivery. In other words, if the ball is going to take the flight path the pitcher wants it to take, the pitcher’s body needs to move in the way he wants it to move. It’s not so much what the delivery is that matters, it’s how consistent the pitcher can be with his mechanics.

We talk about how well pitchers repeat their deliveries all the time–it’s hardly novel to acknowledge the importance of mechanical consistency. Of course, the relationship between this consistency and the particulars of a delivery is not nonexistent–simpler, sounder motions are generally more repeatable than others. There are really two ideas to follow here:

1.) The cleaner the delivery, the higher the “repeatability ceiling.”
2.) Not all pitchers reach their “repeatability ceiling,” and some manage to exceed what seems like their ceiling (usually thanks to athleticism).

Austin Brice clearly has a high “repeatability ceiling”–if we wanted to grade such things on the scouting scale, we might give him a 6, or even higher. When we see him get the delivery right, it looks like he’s close to that ceiling already, and thus has the ability to hit his spots with some consistency. It’s only until you see something go wrong that the clearer picture unfolds.


In his first nine starts of the year, Austin Brice had 31 walks in 39 2/3 innings while striking out 32, good for a 6.35 ERA and a .424 (!) OBP-against. The Marlins were dissatisfied enough to yank him from the Greensboro rotation and move him to the bullpen, where his fastball-curve combination could play up and he could just air it out. Or so you’d think, until Brice faced 49 batters across three outings as a long reliever and walked 14 of them, striking out just seven. He was moved back to the rotation and actually pitched somewhat better the rest of the way (65/33 K/BB in 61 1/3, 5.58 ERA, “only” a .382 OBP allowed).

On June 14, I happened to witness the last of those three relief outings, when Brice came in from the bullpen to relieve Matt Milroy for the sixth inning. The above pitch is the very first thing he threw. It’s obviously wild, but why?

The motion is still basically the same simple, sound delivery from the successful pitches earlier. It’s here that we get a sense of the problem, though–he’s not repeating it.

Brice has a fairly slow tempo to the plate–once he picks his left leg off the ground, it stays in the air for a reasonably extended amount of time. It’s certainly possible to rush a delivery, and I’m not necessarily advocating lightning-speed motions, but the thing about having a slow delivery tempo is that, if things are a tiny bit out of sync, the longer timing can exaggerate the issue. Austin Brice is vulnerable to getting out of sync. At the top of his leg lift, he bends his leg back slightly, and his body follows suit and angles slightly back. Then, as his leg takes a fairly circuitous path to its landing spot, he gets into his arm action and releases the ball just after landing.

It is in the back part of his motion, right at his balance point, that Brice can get out of sync–his upper body (and arm) don’t always stay in line with the lower body. Carry out that slight inconsistency over a slow tempo, and you can get issues like the one in the .gif above, where his upper body fires a bit too late, dragging his arm behind the rest of his body and forcing the ball out late, causing it to make a beeline for Stetson Allie‘s ribcage. Other times, the arm gets into release position before he’s rotated his lower body into position, forcing the ball the other way, like so:


Just watch Brice’s right shoulder in the two .gifs and you can see how different the timing of these two deliveries is. The delivery still looks pretty good here–it’s still easy and loose and all–but he’s not repeating it, and thus the ball goes all over the place. When it is in sync–which, invariably, it is for a fair portion of the pitches in a game–Brice’s stuff plays, but he unravels quickly on the numerous occasions when it isn’t. Out of the windup, where the tempo is slower, Brice’s command is worse (51/44 K/BB) than out of the stretch, where he’s quicker to the plate (60/38 K/BB), which points again toward relief potential, his disastrous June trial in that role and subsequent improvement when re-deployed in the rotation aside. We can speculate if Brice can ever learn to repeat this delivery, or if other pitchers could repeat this delivery, or if he could maybe repeat a different delivery, but the bottom line is he isn’t repeating his motion now, he hasn’t ever done so in his professional career with any consistency, and he’s going to need learn how to repeat a delivery if he’s ever going to capitalize on his considerable potential.

That makes repeating the delivery sound pretty important, doesn’t it? If we’re going to think about command in this fashion, then delivery repetition is a huge key to a pitcher’s ability to establish MLB-caliber command. I could spend all day discussing my perspective on it, but rather than spouting armchair subjectivity, I wanted to test out just how important release point consistency is.


How do we measure release point consistency? There are a number of ways one could claim reflect it, but here’s what I did: I downloaded the raw Pitch F/X charts of the 79 qualified MLB starting pitchers in 2013 from the awesome Baseball Savant database. I then figured out which type of fastball each pitcher used the most, and removed all the other pitches. The reason why I wanted to just look at pitchers’ fastball release points, as opposed to all pitches, is that we often hear talk of another delivery flaw–changing arm slots on different pitch types. This isn’t a good thing, but it really isn’t the same thing as failing to repeat the delivery, because it’s largely intentional. A look at how much a pitcher’s release point varies on his most-used pitch type, though, is going to reflect more clearly on the pitcher’s ability to repeat his motion.

So I calculated the average release point each pitcher used for his most-used fastball (which is, in theory, the release point he’s basically trying to repeat) and then found his average deviation from that release point. It wasn’t the cleanest of processes, so in the interest of full disclosure, I’d like to note two things:

1.) Four pitchers (Doug Fister, Zack Greinke, John Lackey, and Bronson Arroyo) seem to intentionally vary their fastball release points in a way that can’t really be controlled for, so I’m calling them outliers and not considering them, leaving me with a sample of 75 pitchers.
2.) Quite a few pitchers (I would estimate somewhere in the 15-25 range) changed their release point at some point during the season, mostly in terms of where they started on the rubber (and thus the horizontal release point). For those who I identified changes in, I controlled for them by comparing them to the average release point the pitcher was trying for at that time–so if a pitcher switched his release point on May 1, all the April pitches would be compared to the average April release point and all the others would be compared to the average release point from May 1 to the end of the season. It was a judgment call on my part who did make a substantial enough alteration to be controlled for and who didn’t, but I did my best to accurately represent the release point the pitcher was trying for over the course of the season.

So, does this release point consistency data correlate with anything, or is it just totally random? I compared it to a laundry list of control-related variables. Let’s start at the most microscopic level. This is, at its most basic, data about fastballs. How does average fastball release point deviance predict fastball strike rates?


It’s not the most earth-shattering trend in the world, but an r-squared of .15 shows that there’s clearly a relationship–fastball release point variance explains 15% of variance in fastball strike rates. According to this formula, an increase tenth of an inch of average fastball release deviance equates to 3.76% lopped off of a fastball’s strike rate. This is irrespective of the pitcher’s plan or the quality of the fastball in question. To a slightly lesser extent, average fastball release deviance predicts the percentage of the time the fastball is in the strike zone:


I also tested for fastball called strikes and swinging strikes, but there wasn’t a significant correlation there, probably because that has more to do with pitch quality and approach than consistency.

So a pitcher’s ability to repeat his fastball release point exerts some influence over how often his fastball finds the zone and how often it goes for a strike–that seems intuitive enough. But what if we broaden the scope here? How does fastball release point deviance correlate with overall strike rate?


The slope is slightly less pronounced, but the overall correlation is actually stronger than it is for just fastballs. This shows how good of a measure it is at getting to the heart of a pitcher’s command abilities–its predictive value actually increases as one broadens the scope of the data. The reason that it predicts overall strike rate better than fastball strike rate has to do with, in my belief, the different roles pitchers use fastballs for. Every pitcher’s overall goal is the same–don’t allow baserunners, which involves throwing strikes to some degree–but pitchers may have different goals for their fastballs. For R.A. Dickey, it’s a change-of-pace pitch. For Bartolo Colon and Cliff Lee, it’s a pitch to pound the zone with. For Madison Bumgarner, it’s a swing-and-miss offering. For Mark Buehrle, it’s a Livan-esque pitch to put just on or off the corners and induce weak contact. Ricky Nolasco only put his fastball in the zone 36.15% of the time and got a league-worst 56.7% strike rate on it, but that’s because he’s very cautious with it, not because he has no idea where it’s going. His overall numbers (43.1% zone, 63.89% strike) are much more indicative of his command ability. There’s not as strong a correlation with overall zone percentage:


But interestingly, there is a correlation with first-pitch strike percentage:


Of course, all these correlations are interesting, but let’s take this out of the realm of mere pitch-by-pitch data and work with the most common arbiter of control: walk rate.


The metric still holds up as having predictive value–as much predictive value as it has for fastball strike rate, in fact. According to this formula, .05 inches of fastball release point deviance predicts an extra 1% on a pitcher’s walk rate. You can see that every pitcher with a walk rate below 5% is around average or above in release consistency, whereas all those with rates above 9% are around average or worse. It even holds a sliver of predictability on strikeout-to-walk ratio:


While fastball release point consistency doesn’t explain everything about pitching, it does appear to have a surprisingly robust influence on a number of control-related factors. Certainly, much more study is needed to figure out exactly how powerful of a mechanism it is in a statistical sense. With many prospects getting Pitch F/X looks between Spring Training and the AFL, a look at their fastball release point deviations may shed some light on how they project in the control realm, in addition to their minor league numbers and raw mechanical profiles. I don’t know what Austin Brice‘s average fastball release point deviance is, but I’m guessing it’s above the .226″ average in this sample, and if he can get to that realm of consistency, he just might unlock his potential.

Print This Post

Nathaniel Stoltz is a prospect writer for FanGraphs. A resident of Bowie, MD and University of Maryland graduate student, he frequently views prospects in the Carolina and South Atlantic Leagues. He can be followed on Twitter at @stoltz_baseball.

Sort by:   newest | oldest | most voted
Jon Roegele

Wow another long read Nathaniel!

I wanted to mention that using PITCHf/x release points is a little tricky, as you probably know, in that what you’re calling “release points” are really the ball location that the fitting algorithm projects at 50 feet from home plate. The actual average release point for MLB fastballs is close to 54 feet, based on Trackman summaries that I’ve seen. So a taller pitcher (or one who extends more than average) would have the advantage of releasing the ball closer to the 50 foot point where your numbers would be from. At least I don’t think the Baseball Savant numbers are adjusted back to 55 feet or anything.

Also, these numbers are going to look different park-to-park, so I suspect a starter who happened to pitch more at home than away may have an advantage from a “release point consistency” perspective. And yes moving along the rubber, even slightly, can affect the numbers, and this isn’t always easy to spot (I don’t think, anyway).

That said, I would expect there to be a pretty high correlation of consistency at 50 feet than at the actual release point, these are just some issues that may be affecting your correlation plots.

Interesting work!