Pitch speed and balls in play

Ever since the rules of the game were changed and pitchers were allowed to throw overhand, fans and baseball men alike have been drawn to flamethrowers. We marvel at the sound of a fastball whizzing through the air, the smack as it hits the glove. We debate whether Walter Johnson threw harder than Bob Feller or Nolan Ryan. A pitcher like Steve Dalkowski can live on in memory despite not being able to hit the broad side of a barn, because he’s capable of throwing the ball through it.

We know that strikeout rate is one of the key indicators of a pitcher’s long-term success—largely because strikeouts keep the ball out of play and all the bad things that can happen there. And we know that a pitcher who throws hard is more likely to strike out a lot of batters than one whose top speed is “batting practice.”

But is throwing hard beneficial to a pitcher even when the ball is put into play?

Let’s take every pitcher with more than 100 balls in play in either 2007 and 2008 and figure out his average pitch speed, BABIP, and strikeout rate (expressed as strikeouts per batter faced). Many pitchers appeared in the data set for both years, although some only made it in one year.

I went with average pitch speed instead of the more common average fastball speed because I thought that including the mix of fastballs and offspeed pitches gave a better approximation of how hard the pitcher was throwing.

Next, we divide the pitchers into four categories based on the average pitch speed: less than 85 mph, between 85 and 87.5 mph, between 87.5 and 90 mph, and above 90 mph. Grouping the pitchers is intended to smooth out other factors that we know affect BABIP—things like defense, walk rate and home run rate. Granted, it’s not a perfect solution, since it’s plausible a few teams could have a disproportionate amount of pitchers in one category, which might introduce the effects of defense into the mix. Also, average pitch speed may be correlated to walk rate or home run rate, which would then incorporate them into the equation.

And of course we need to watch out for the possibility of selection bias. By choosing a lower limit of 100 balls in play for our sample, we run the risk that, for example, 90 percent of the pitchers who didn’t make the cut threw 90 or above and allowed a BABIP of .400. That’s not a very likely scenario, but there’s definitely a chance of a smaller effect. Since we’re not being too precise in this investigation, we can continue unmolested.

We proceed using our groupings, and find the following samples.

For 2007:

Speed       n           BIP
85-         80          28509
85-87.5    108          33019  
87.5-90    100          33214
90+         64          15847

And in 2008:

Speed       n           BIP
85-         94          30031
85-87.5    136          39291  
87.5-90    128          39316
90+         64          15663

We anticipate that the strikeout rate will go up as pitch speed increases. We all expect Randy Johnson to strike out more batters than Jamie Moyer.

The data for both 2007 and 2008 confirm our impression. Strike out rate does increase along with average pitch speed.



Why is this important? Voros McCracken and J.C. Bradbury have already demonstrated a correlation between strikeout rate and BABIP. We see the same thing with pitch speed in 2007, but 2008 fails to follow the pattern. My guess is that’s just a blip in the data, since pitch speed and strikeout rate appear to track pretty closely, but I’m not entirely certain.


Using Recurrent Neural Networks to Predict Player Performance
Technology is rapidly advancing possibilities in decision-making.


So we’ve got three variables all correlated: BABIP, strikeout rate and pitch speed. But as every good statistician knows, correlation does not equal causation. Can we infer some more meaningful relationship here?

To try to tease a little more meaning out, let’s bring in batted ball types for our pitchers. Here’s the breakdown of the types of balls in play for each category of pitch speed.



We can see that pitchers who throw harder tend to give up more ground balls (the green areas in the graphs). And previous studies have shown that ground ball pitchers tend to have a higher BABIP than fly ball pitchers, although they tend to gain back most of the advantage on slugging percentage in play. This makes the relationship we see here even more surprising.

Based on this additional information, there’s a couple of ways our causation arrow could point. Higher pitch speeds lead to more strikeouts, which leads to the batter being defensive late in the count, and therefore more apt to hit a weak ground ball. In a similar vein, faster pitches give the batter less time to react, so he’s more likely to generate weak contact, and either hit a ground ball or foul it off.

Alternatively, it’s possible that the apparent relationship between pitch speed and ground ball rate is more illusory than illuminating. Both could independently affect a pitcher’s BABIP rate without one causing the other. This feels less likely however, since a causal relationship just seems like an obvious fact (although that line of reasoning has bitten us in the past).

While there’s no conclusive proof that how hard a pitcher throws plays a role in his BABIP performance, there are at least a few wisps of smoke that make this an avenue worth exploring.

Further investigation is needed to determine whether pitch speed truly does have an effect on BABIP above and beyond its contribution to strikeout rate, perhaps in the area of ground ball ratio. Other areas for inquiry include looking at slugging percentage in play, both batting average and slugging percentage on contact (which includes home runs), and maybe devising a better way of measuring how hard a pitcher throws than simple average pitch speed.

Print This Post

Comments are closed.