Dialing it up

For a while now, I’ve been wanting to do some digging into how the ball-strike count affects a pitcher’s velocity. I’ve always assumed that velocity correlates inversely with run expectancy—for example, I figured that the most extreme “pitcher’s count,” 0-2, would yield the highest velocities since a pitcher is trying to power a pitch by the hitter for a strikeout and has the highest margin for error of any count.

Also, I found it reasonable to assume that on a 3-0 count, a pitcher can loosen up a bit in order to avoid walking the batter. I finally did some number-crunching on this topic, and it turns out that my intuitions reflected reality.

Before I get into any of the data, there’s one key thing I need to mention: I’ll be looking exclusively at fastballs in this article. There are two reasons for this.

One is that fastballs would appear to be the least susceptible to classification errors; I think offspeed pitches are more likely to be misclassified due to the similarity in velocity. Although I’ve done classifications for a lot of 2010 pitches, there are always going to be some errors. For pitches that I haven’t classified, I use the MLBAM classifications.

The other reason is that pitchers don’t tend to throw offspeed pitches in pitchers’ counts (particularly 3-0), which makes it difficult to get information for all counts.

Establishing some baselines

In order to best understand the differences between how pitchers approach different counts, we first have to set up some league averages. The table below shows a pitch’s velocity difference in each count compared to its overall velocity. These data are limited to a pitcher’s primary fastball (four-seamers for most guys, sinkers for guys like Derek Lowe, cutters for a handful like Mariano Rivera) if it has been thrown in each of the 12 ball/strike permutations.

A few pitchers who threw an offspeed pitch more commonly than a fastball were eliminated from the sample. Also, keep in mind the classification issues I mentioned earlier. So basically, take the exact numbers with a grain of salt, but it should give us a pretty good idea of league-wide trends. The table is sorted by increasing run expectancy.

0-2 1-2 0-1 2-2 1-1 0-0 2-1 1-0 3-2 2-0 3-1 3-0
Difference +0.56 +0.62 +0.06 +0.62 +0.02 -0.27 +0.03 -0.31 +0.48 -0.32 -0.14 -0.58

As you can see, a pitcher typically throws his hardest fastball in two-strike counts. I was incorrect in my assumption that 0-2 counts are hardest on average, but they’re close. As I expected, 3-0 counts generate the slowest velocities.

How can we use this to identify pitchers as “consistent” or “inconsistent” with their velocities? What I’ve decided to do is take the absolute value of the distance between a pitch’s average velocity and its velocity in each count and weight it by the number of times it was thrown in each count. To get a feel for what this looks like across the league, take a look at this distribution (using fastballs with at least 10 pitches in each count) below.

image
Most pitchers have an average difference of around 0.3 mph. According to my data, no pitcher was lower than 0.1, and only one was higher than 1.0. I’ll raise the ol’ skeptical eye at some of these guys whose pitches I haven’t classified really well.

Looking at specific pitchers

Of the 257 pitch types, here are the pitches that have fairly “consistent” velocities:

Pitcher Pitch Type Fastest Count Slowest Count Avg. Velocity Difference
Jaime Garcia Four-Seam 2-2 3-0 0.11
Jamie Moyer Sinker 2-1 0-1 0.11
Derek Lowe Sinker 3-2 3-0 0.12
Brian Tallet Sinker 1-2 2-1 0.13
John Lannan Sinker 0-2 3-0 0.13
John Ely Four-Seam 1-1 3-0 0.14
Pedro Feliciano Sinker 1-2 1-0 0.14
Esmil Rogers Four-Seam 3-2 3-0 0.15
Jason Hammel Four-Seam 0-2 2-0 0.15
Jeanmar Gomez Sinker 0-2 3-0 0.16

And those with velocities that are apparently affected by the count:

Pitcher Pitch Type Fastest Count Slowest Count Avg. Velocity Difference
Carlos Zambrano Four-Seam 3-2 3-0 1.01
Brett Anderson Sinker 3-2 3-0 0.97
Robinson Tejeda Four-Seam 0-2 2-0 0.97
Anibal Sanchez Four-Seam 0-2 3-0 0.96
Rich Harden Four-Seam 0-2 0-0 0.94
Clayton Richard Four-Seam 0-2 3-0 0.86
Travis Wood Four-Seam 3-2 3-0 0.82
Javier Vazquez Four-Seam 0-2 3-0 0.78
Brandon Morrow Sinker 0-2 2-2 0.78
Brandon Morrow Four-Seam 0-2 1-0 0.76

A few names here I’m extremely skeptical of due to classification issues—Zambrano and Tallet look questionable to me. I know that Zambrano has a wide range of fastball velocities, but he also throws a cutter, sinker and splitter, and there are surely some misclassifications. Tallet’s two-seam group has a few others thrown in.

Also, something strange is going on with Morrow’s two-seamer. That pitch and his hard changeup are pretty close in speed, but I went game-by-game and split them apart. Interestingly, his two-seam velocity on 0-2 counts was 96.1 and only 91.0 on 2-2 counts. Morrow’s a guy whom I’ve seen range from the mid-80s to the high-90s within a single game, so my assumption is that he threw more 2-2 two-seamers in his “slow” innings.

A Hardball Times Update
Goodbye for now.

From what I know about them, I’m not surprised at all to see Anderson, Tejeda, Richard or Harden on this list. Justin Masterson would be high up (0.82) if he had enough qualifying pitches. A few weeks ago, I covered Masterson’s repertoire and mentioned the big range he gets on his velocity, so it’s nice to see that these numbers back that up.

Doing more with this concept

The point of this article was, more or less, to establish some general trends and not buy too much into the exact numbers, given the uncertainties I have concerning pitch classifications. For pitchers I’m familiar with, it’s pretty cool to look at the numbers here and see where they fit in league-wide.

For example, Joba Chamberlain is someone whom I’ve watched a lot and I recall having a lot of range on his velocity. I also remember Zack Greinke being the same way. The four-seamers of Greinke and Chamberlain are 29th and 30th, respectively.

A.J. Burnett and Phil Hughes are two Yankee starters whom I remember being pretty consistent; Burnett is 68th most consistent, and Hughes is 91st, both substantially (but not insanely) more consistent than the average pitcher.

That’s all I have for now, but I would like to return to this topic in the future. Here are some things I’m thinking about:
{exp:list_maker}Figuring out a reasonable way to look at velocity range for offspeed pitches.
Improving pitch classifications to increase accuracy.
Tying in velocity change within a game in addition to count in order to look at more of an overall picture for velocity variance. Jeremy Greenhouse has done some good work on this subject. {/exp:list_maker}

References & Resources
PITCHf/x data are from MLB Advanced Media and are here courtesy of Joe Lefkowitz’s tool. Pitch classifications, when available, are by the author; otherwise, they are from the MLBAM algorithm.


5 Comments
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Lucas Apostoleris
12 years ago

That part is fixed now.  Thanks for pointing it out, Dave.

garik16
12 years ago

Not a particularly large difference really, which makes a good bit of sense. 

Minor point:  The Classification system is best seemingly at classifying certain breaking balls (curveballs particularly)…..I find it can’t tell the difference between FF and FT way too often, which can of course have different speeds.

But for sample size reasons you definitely wanted to go with fastballs.

Dave Studeman
12 years ago

Love the trend, Lucas.  Makes a lot of sense.

But I don’t understand your classification system. When you say this…

take the distance between a pitch’s average velocity and its velocity in each count and weight it by the number of times it was thrown in each count.

…what do you mean? I just don’t understand this, and the accompanying graph doesn’t make sense either.  Why would all pitches be above average in velocity?  Shouldn’t the difference between velocity and average velocity come out to zero?

Dave Studeman
12 years ago

One other thing.  When you say “257 pitch types,” do you mean 257 pitcher/pitch type combinations?

Lucas Apostoleris
12 years ago

Dave, yes.  The 257 pitch types means two different pitcher and pitch type combinations.  Like, Brandon Morrow’s four-seamer and sinker are counted as two different pitch types.

As for your first question … I realized I made the error of not saying that it’s the *absolute value of the distance* of a pitch’s average velocity and its velocity in each count.  I hope that makes it more clear; sorry about that.