About a year and a half ago, Baseball Prospectus revealed a suite of catching stats that formed the basis for our industry-leading valuation of catchers. These new stats would shape how we perceived and discussed catcher value, but they also opened the door to better understanding the performance of pitchers.

Two key statistics—CSAA and CS Prob—serve as the basis for the pitch framing portion of our catching metrics. Today, we’ll show how those same statistics can tell us a great deal about pitching as well. CS Prob was initially introduced in 2014 with Harry Pavlidis and Dan Brooks’ first catcher framing model. Early the next year, Jonathan Judge joined the effort and the team introduced CSAA, officially moving our framing models beyond WOWY.

Of the two, CS Prob—short for Called Strike Probability—is the more straightforward: the likelihood of a given pitch being a strike. CS Prob goes beyond what the strike zone ought to be and instead reflects what it is: a set of probabilities that depends on batter and pitcher handedness, pitch location, pitch type, and count. Good pitchers understand that while the strike zone is a dynamic construct, it nonetheless has some consistencies depending on which combinations of these factors are present. We calculate CS Prob for every pitch regardless of the eventual outcome.

The other statistic, CSAA, stands for Called Strikes Above Average; a measure of how many called strikes the player in question creates for his team. In the case of catchers, we isolate the effects of the pitcher, umpire, and other situational factors which allows us to identify how many additional called strikes the catcher is generating, above or below average. For catchers, this skill is commonly described as “framing” or, in more polite company, “presentation.”

For pitchers, we can apply a similar methodology—controlling for the catcher, umpire, etc. to identify the additional called strikes created by the pitcher. CSAA is calculated only on taken pitches, an important nuance. A pitch must be taken in order to be eligible to be called a strike by the umpire, so while CS Prob looks at all pitches, CSAA only takes into account pitches where the outcome is left up to the umpire.

What can these two statistics tell us about pitcher performance and skill? First, we should define a few important things:

Control – The ability to keep the ball in the strike zone, though not necessarily in any particular location within that zone.

Command – The ability to precisely locate pitches, in or out of the zone, with the goal of keeping each pitch out of the heart of the plate.


At its core, CS Prob tells us the likelihood that each pitch is going to be called a strike. Having a high CS Prob is an indication that the pitcher in question is throwing a lot of strikes and largely keeping the ball around the strike zone. CS Prob fundamentally tells us which pitchers pound the strike zone, regardless of the quality of those offerings.

Take for example this graphic that showcases, from the catcher’s perspective, some rough zones for called strike probabilities for a right-handed pitcher facing a right-handed hitter in an 0-0 count:

For the 2016 season, the inner most circle represents pitches that are called strikes 90 percent of the time. The larger second circle, which stretches from the top of the strike zone to below the bottom of it, includes pitches that are called a strike roughly half the time. Finally, the third largest circle, which stretches almost entirely outside the zone itself, represents pitches that go for a strike about 10 percent of the time. The reference strike zone is reflected by the black square, allowing you to see the difference between where the strike zone was supposed to be versus where it ended up being called.

These zones change and shift based on the count, handedness of the batter or pitcher, and even the pitch in question. Overall though, pitchers who have a high average CS Prob are clearly working within the confines of the strike zone.

There is a clear and obvious connection between the concept of control and CS Prob. Think of pitchers who pound the strike zone. Among pitchers with at least 100 innings, Bartolo Colon led the league with a 52.1 percent CS Prob. That is, any given pitch thrown by Colon has more than a 50 percent chance of being called a strike. The table below showcases the top 10 pitchers in the major leagues for 2016:


CS Prob

Bartolo Colon


Rich Hill


Jimmy Nelson


Steven Matz


James Paxton


Mike Foltynewicz


Hisashi Iwakuma


Max Scherzer


Shelby Miller


Clayton Kershaw


CS Prob is an important piece particularly as distinguished from the rulebook strike zone. Take the CS Prob champion, Bartolo Colon: during the 2016 season, Colon lead the majors with a 52 percent CS Prob. As we’ve established, that means that more than five out of every 10 pitches from Colon is likely to be called a strike. If Colon were subject to robot umps calling the rulebook strike zone, his Zone% tells us he would get strikes on 50 percent of his pitches. That means Colon is getting two extra strikes per 100 pitches simply because the strike zone that is called isn’t the one the rulebook lays out. Rich Hill on the other hand actually gets penalized, as his Zone% of 52.6 percent is higher than his CS Prob.

Generally speaking, CS Prob and Zone% should correspond with each other because the called strike zone is generally pretty close to what the rulebook defines as the zone. Still, a few percentage points one way or another leads to a few more or a few less strikes over the course of a game. That could mean a few dozen strikes going the other way over the course of a season, and the value of a strike is sometimes really significant.

Throwing pitches in the strike zone as defined by the rulebook isn’t necessarily a ticket to success; much like real estate it’s all about location, location, location.


Now that we’ve established that CS Prob is a proxy for control, we can build on it. After extensive review, we’ve concluded that CSAA substantially reflects a pitcher’s ability to command his pitches. It’s important to make the connection between what CSAA does and the popular definition of command.

Traditionally command is understood as the ability to “hit your spots”—having the ball end up where you intend it to. Over the years this has been studied in numerous ways—most notably by attempting to determine how much the catcher moves his glove to receive a pitch. This is flawed because the catcher’s glove isn’t always the target, and we can’t know where the pitcher is truly intending the pitch to go.

What we can do is come at command from a different angle. A pitcher with good command should be more predictable for the catcher—their pitches often end up in the locations, and with the movement that the catcher expects. This skill results in easier receiving for catchers, and additional called strikes for the pitcher. Once we aggregate the data cross thousands of pitches, CSAA is able to tell us whether a pitcher is reliably hitting his spots.

CS Prob is actually covariate in the model for CSAA, which is a fancy way of saying that CSAA measures the extent to which a participant tends to affect the likelihood of a strike being called, notwithstanding its final location. As such, CSAA controls for all of the same things as CS Prob and adds in the umpire and catcher for good measure.

So what does accumulating CSAA look like? It’s not as easy as it sounds. Sure, you could throw a ton of pitches in the middle of the zone and basically guarantee that you’ll wrack up called strikes on the pitches hitters don’t offer at. The downside to that approach is that pitches in the center of the plate get crushed.

Your browser does not support iframes.

The best command pitchers actually have a somewhat interesting approach. They tend to work further outside the edge of the strike zone than you might think, trying to pick up extra strikes in the 20 percent or 30 percent bands for CS Prob. When working within the zone, they avoid the center of the plate—the 90 percent zone—and focus on something in the 75 percent range. The very best command pitchers—guys who are one standard deviation or more above the mean—also have a high propensity to throw pitches in the sub-10 percent band as well. When these guys are trying to work out of the zone, they really work out of the zone. That means spiked curveballs, down-and-out sliders, and eye-high fastballs.

You can see in the chart below the density of offerings between the best and worst command pitchers and where they tend to concentrate their pitches along the spectrum of called strike probability:

You’ll notice that most of the positive values are located in the bottom half of the CS Prob range: 0 percent to 50 percent. This means that the best command pitchers actually work outside the effective strike zone—especially in the 0 percent to 9.99 percent range—as compared to the worst CSAA performers. Simply put: if you have good command, you don’t have to throw as many strikes.

Command thus differs from control because pounding the zone doesn’t necessarily mean that you’re hitting your spots. A guy like Clayton Kershaw can get away with having a high CS Prob (99th percentile) and an all right CSAA (69th percentile) because his raw stuff is so impressive. For someone like Colon, however, elite command (top six percent in CSAA) is critical to his success. Luis Perdomo combines throwing a lot of strikes with poor command (59th percentile CSAA) to disastrous results—nearly 1.5 home runs per nine innings.

As such, command is something that you don’t need, provided you are blessed with dynamic stuff. For most pitchers, though, they need to work the edges of the zone effectively, gathering up extra strikes as much as possible in order to be successful.

The best command/CSAA pitchers tend to be guys who produce good results despite seemingly lackluster or otherwise inconsistent stuff. The top 10 pitchers with a minimum 500 chances in 2016 are listed below (note: percentages here indicate how many more strikes the pitcher got than CS Prob would otherwise indicate).



Zach Davies 3.5%
Josh Tomlin 2.8%
Kyle Hendricks 2.5%
Ryan Vogelsong 2.5%
Mike Leake 2.3%
Zack Greinke 2.1%
Wily Peralta 2.0%
A.J. Ramos 1.9%
Jon Lester 1.9%
Chris Young 1.9%

A pitcher like Davies—a righty whose fastball barely broke the 90 mph barrier—needs to work those edges in order to be successful at the highest level of baseball. Throughout much of his career as a prospect Davies seemed destined to be a Quad-A player because of his unimpressive stuff, but the Brewers are now reaping the rewards of Davies’ command: his pitches are 3.5 percent more likely to be called a strike than the average pitcher.

Command can be a powerful tool. For Davies, it converts a seemingly fringe prospect into a three-win pitcher. A decline in command also corresponds to the difference between Jered Weaver being a league-average pitcher a few years ago to whatever it is he was in 2016. Take a look at how his CSAA and WARP have changed over the past three seasons:

Year CS Prob CS Prob Rank CSAA MLB Rank Max FB Velo HR/9 WARP
2014 65.3% 73rd 2.1% 2nd 89.7 1.1 0.4
2015 66.2% 12th 1.5% 7th 87.2 1.4 -1.7
2016 62.7% 33rd 0.6% 57th 86.2 1.9 -5.3

Suffice it to say that the margin for error for a pitcher with a mid-80s fastball is razor thin—maybe even less than 1.5 percentage points, if the table above is any indication. It’s not enough to simply look at Weaver’s declining CSAA and say that losing an extra strike or two per game hurt him so badly. His declining CSAA points to overall diminishing command. The result? More pitches over the middle and Weaver’s home run rate has soared, while his value has plunged in the other direction.


Don’t just take our word for it. Historical CSAA and CS Prob stats are now included on our stats pages, so you too can pull data for pitchers like Weaver and assess the impact of improving or declining command on their careers.

There is one special case that’s worth pointing out and discussing in some length. Tom Glavine’s 2008 season doesn’t seem overly special when you look at raw CSAA. His 5.3 percent mark is certainly impressive, but ranks just 120th among all seasons since 1988.

Where Glavine’s season really shines is when you control for the era and seasonal factors. Creating Z-scores allows us to compare each pitcher to their peers, helping us better understand how they are performing compared to their peers that season. This is important because as PITCHf/x began being utilized by major-league teams and the league office, umpires started calling games differently. Specifically, they got a lot better and more consistent.

That’s why Glavine’s 2008 season is so remarkable. His 5.3 percent CSAA is barely half of Greg Maddux’s best-ever mark, but that 5.3 percent was more than seven (!!!) standard deviations better than the mean for the 2008 season. By 2008 umpires had really improved in terms of calling the strike zone, but Glavine was still working like it was 1995.

We would also be remiss if we didn’t mention the sheer dominance of Maddux, who from 1995 to 1997 was getting a strike bonus of 8–10 percent on every pitch he threw. That is, put bluntly, absurd.


There are of course two ways to analyze and understand pitcher performance. Let’s use pitch movement as an example. We can look at PITCHf/x data and see that a pitch dropped 10 inches horizontally and traveled at 73.7 mph. We can also look and see that against that particular pitch opposing hitters swung and missed 15.4 percent of the time, hit ground balls 10.7 percent of the time, resulted in a strike (either via a called strike, a whiff, or a foul ball) 26.6 percent of the time.

For the pitcher, the measurements are a means to an end. The most important thing is that the pitch delivers positive outcomes and/or helps set up other pitches to have positive outcomes. Having the best stuff in baseball doesn’t mean anything if you can’t get hitters out.

Our approach to understanding command and control is rooted in the latter, as we look to describe the outcomes and intent related to pitch location. We’re able to do this by using CS Prob and CSAA to isolate the performance of the pitcher from all other influencing factors, and to establish an idea of how an ability to locate the ball—either in the zone, or precisely outside of the zone if need be—impacts the intended outcome. This is unlike COMMANDf/x, a methodology that uses glove movement to measure precision. Catcher movement can be helpful, but it can also be misleading.

Being able to look at command and control through the lens of strikes is critical, because that is the currency in which pitchers trade throughout an at-bat. With CSAA and CS Prob, we believe that we’ve identified the tools to do just that.

You need to be logged in to comment. Login or Subscribe
Two thoughts: (1) This is excellent research IMHO - carefully conceived and executed, and clearly presented. Well done and many thanks to all involved. (2) I knew Jered Weaver had a rough year last year, but that -5.3 WARP figure is breathtaking.
DRA absolutely hates him. As we work on offseason updates, it's certainly front of mind. That said, our accuracy on how bad a pitcher really is who is indisputably bad is not exactly the zone most people / teams are interested in.
tomorrow we'll be able to see how much he relies on changing speeds, too. At least that part he's kept up, but ... not to much success.
Is there a way to export to CSV all seasons from the sortable stats page instead of year by year?
OK, I admit that this has been bugging me for years and I have to say something. Command is just really good control so let's quit pretending they are two different things. The inside baseball that alleges some kind of nuance between the two is a myth, gone unchallenged for years. Look at the charts. A pitcher capable of throwing a ball within a 2" spot has excellent control and his chart looks like the "command" chart. One who can locate within 6" throws strikes, but his chart looks like the "control" chart. One who can't do even that has poor control and walks a bunch of guys. The entire command vs. control argument is absurd. If you need further proof, look over prospect and scouting reports which consistently confuse the two terms, frequently using command when they probably mean control, at least in the context of the accepted definitions. I do believe that at some point in the past, command and control did have distinct definitions, but that is no longer true. To me, control is being able to put the ball WHERE you want it. Command is more subtle, the ability to make a pitcher DO want you want. Get that curve to break a little sooner, a little later, a little sharper. Can you reliably get your fastball to move in over (or out away) the outside corner? Changing speeds without losing pitch quality is an element of command. Being able to get your pitches to perform like that on cue, THAT is command and it has f-all to do with control, i.e. where you throw the pitch instead of how. But do you ever hear it described that way? It is always the ridiculous definition given here, that control is throwing over the plate, command is throwing on the corners. That's like trying to argue "doubles power" and "HR power" are two different skills. No, they aren't. HR power is just above-average doubles power. Another way to put it, you can't have HR power without doubles power, too. Similarly, you can't have command as defined here without control, they are the same skill at different points on the same scale.
I think there is a lot of thinking to be done, hopefully much of it publicly, about the relationship between control and command. Most people have indeed assumed that control and command are inextricably linked, including us, but I do think that's an assumption worth prodding at a bit. If we are going to treat control as being about ability to hit the strike zone, though, the reality is that there are some pitchers (nibblers and others) who try really hard to work around the strike zone and succeed best when they do so. As such, they could exhibit quality command and yet not be in the zone as often as others. I'll also say that command is, for us, almost certainly more important and will correlate better with end results. As the article notes, you can have a 100% CS_Prob in theory, and your RA9 will probably be 100 as well if those baseballs ever come back down to earth. CS_Prob seems to be where it starts, and the question is what you do from there.
I have to say that they are not just linked, but one and the same. I also want to point out that the only way to get the control pattern in the chart is for a pitcher with a margin of error +/- 8" and who is aiming for the middle of the plate with each pitch. He'll throw close to 100% strikes this way, but probably get hammered as you noted on the relatively high percentage that actually cross the middle of the zone instead of "missing" and finishing on an edge. On the other hand, if he aims for an edge, approximately 50% of his pitches will be balls. Aim for an upper or lower corner and strike rate drops to 25%. Consider the same pitcher with the accuracy noted in the command chart. His MOE is +/- 2". Aim for an edge, his strike rate is still nearly 100% because the ball is 1.5" wide. Aim for a corner and the strike rate is still probably 80% or so. The only difference between those two pitchers is degree of control. First guy is average, second guy has "pinpoint". What is described as control and command is just where they target the pitch, unless some pitchers can be very precise when they throw to the middle of the plate, but their grouping suddenly expands when they try for the edge of the zone. I'm not sure I buy that as a common phenomenon. I concede that even in my definition (and it is mine alone, I didn't mean for it to be a general truth), control and command are not completely independent. It would be unusual for a pitcher to excel at one, but completely fail at the other. Still, it isn't out of the question that a hurler might have the ability to vary the break and speed on his slider at will, but struggle to locate it properly. He has command of the pitch, but no control. Then you have a pitcher who reliably locates his fastball on the outside edge all day long, but struggles with location should he try to take something off or attempt to throw to the inside corner instead. Thanks for taking the time to reply. Despite any gripes I might have, the article successfully forced me to rethink the topic and its implications.
Very interesting article. This hasn't been bugging me for years, but last week I decided that what jnossal has stated is true - control may indicate the ability to throw strikes, but command as it is currently defined is just better control. However, I think CSAA may be a way at actually getting at command. I really would like a better description though of how it is measured. First you say that looking at catcher movement is flawed, but then in the next paragraph it sounds like you are measuring catcher movement ? Another aspect of command that CSAA misses is the issue that you are getting at with your graphic of best-worst - and I think this is one of jnossal's points: Pitchers with command are getting pitches to move reliably, whether to the edges of the strike zone or darting out of it. This would seem to be part of command, and then whiffs outside of the strike zone should also be part of its calculation. (Envision Andrew Miller in the playoffs, or vintage Mariano) (Although this might be problematic because overpowering fastballs might be counted here as well).
solely relying on the assumption that 'glove is target' is one thing, making the assumption that a 'more catchable pitch' is another. We've done the latter, while commandfx relies on the former. There is much more to command than this aspect, for certain. For example, can you put your curveball back-to-back in the zone as a drop-in followed by out of the zone in the dirt?
Off the top of my head, if I had a choice between command f/x with all of its weaknesses, and CSAA (which is much "cleaner"), for identifying and quantifying "command" I think I would much prefer the former. "Cleaner and more precise" doesn't necessarily mean you're capturing something more accurately than some messy or brute force type of metric like command X. Then again I don't think they are exactly capturing the same thing, so the results of one or the other might be more or less useful for one inquiry or another.
Does CSAA for pitchers control for umpire and catcher? Throughout the article you keep saying CS prob is "called strikes per pitch." As in: "That is, any given pitch thrown by Colon has more than a 50 percent chance of being called a strike." "That means Colon is getting two extra strikes PER 100 PITCHES simply because the strike zone that is called isn’t the one the rulebook lays out." CS Prob is percentage of strikes CALLED per CALLED pitch right? That being said one of the difficult issues with this model and analysis is that the quality of a pitcher's stuff affects whether a pitch is swung at or not and thus the pitches not swung at are a selected sample of pitches that are inherently related to stuff. I'm sure you guys are aware of this. For example, for a pitcher that has nasty stuff, the pitches that are not swung at may tend to be very much away from the K zone giving the impression that the pitcher has poor control even if he doesn't. If a pitcher does not have good stuff batters may be able to lay off pitches just outside the zone giving the impression that the pitcher has better control than he does have and perhaps better command (IDK about that). You might want to consider ALL pitches for a control metric. If fact, I'm not sure why you wouldn't. Why wouldn't you? As far as the definitions and differences between of command and control, YOU can define them anyway you want. There are no parochial definitions for those words. They general way the authors have chosen to define them is useful and reasonable and exactly the way I like to define them. Command being simply how close a pitcher can come to his intended location (which tends to vary a lot with pitch type even within pitchers). Control is simply how many pitches end up in the zone. Control is generally a function of intention and command ability. If I have poor command and great stuff I am going to just fire balls near the center of the zone and have good control. If I have poor command and poor stuff I am going to be forced to stay away from the middle of the zone and because of my bad command my control will be awful (and I'm probably an awful pitcher). You can go through all the possible iterations of "stuff" and "command" and you will arrive at a pretty good projection for control. But again, if you're measuring control (and perhaps command) using only called pitches you are going to run into some trouble.
Thanks for the thoughts MGL. CS Prob is calculated for all pitches, so that should remove some of the noise in terms of guys with great stuff getting a lot of swings at pitches in the zone. We also agree that any measurements of control (and perhaps command) are going to intrinsically include some aspect of approach, in agreement with the example you gave. CSAA on the other hand is only calculated on called pitches, as we can't know what the umpire would have done if a pitch had not been contacted or swung at. We do control for pitcher and umpire with CSAA as part of the model, so this is the output that is distinct to the pitcher controlling for those factors, among others.
Thanks for the clarification (or more accurately, correcting my misunderstanding - your definition of CS Prob was clear) and your other comments!
I am not at all convinced how much of your definition of command is captured in pitcher CSAA. Certainly some, but it is not clear, to me at least, how much and to what extent what it does capture is useful. I do like the analysis and methodology though.
A few immediate reactions, of which at least some I assume get addressed in upcoming pieces: * How quickly does pitcher CSAA converge, and how stable is it over time? * How does pitcher CSAA correlate with other measures of pitching ability? * Do you see any categorical differences in pitcher CSAA by, say, handedness, velocity, or repertoire? * In particular, how do knuckleballers fare? * I too am concerned about the lack of data from pitches at which the batter swings, and the uneven nature of their exclusion from pitcher to pitcher, but don't have any good ideas for how to remedy this problem. * CSAA as an acronym still makes me think of Caught Stealing Above Average. Can you choose a different name? :)
great points! 1. off the top of my head, it's almost as fast (# of pitches) as it is for catchers. So a couple weeks of games is a reasonable rule of thumb, but we should dig out the proper values 2. we'll have more on that this week 3. looks like harder throwers/nastier breakers fare poorer, even though pitch type is a factor the specifics (movement speed etc) are not 4. they really create their own special pool, don't they? Especially since we're being specific to pitch type. In general, they're hard to frame 5. me too. That's why we have more stuff in the pipeline, some of which builds on the tunnel/sequence data we just published today 6. stat acronyms will be the death of me
Thanks, Harry. It certainly makes sense that better stuff is both harder to catch and harder for umpires to judge accurately. I don't know whether that better stuff also correlates (negatively) well with a pitcher's ability to place a pitch near the intended target, but the reduction in called strikes is notable on its own. Possibly I was not clear when I asked about stability. I was trying to contrast with convergence and get at how likely it is to change quickly. Are there big jumps month-to-month? CSAA is also the California State Automobile Association, but that's not so common. I suppose you could do Called Balls Below Average instead, but that's a bit counterintuitive.