EDIT: I am in the process of working out some data corrections to the PITCHf/x data, and I have updated this post with corrected pitch speed data.
Dave Cameron wrote a piece yesterday at Fangraphs about Justin Verlander’s fastball speed (hat tip to Tango). I love both Dave Cameron’s work and Fangraphs. Fangraphs is quickly becoming one of my very favorite sites on the InterWebs. However, something about Dave’s post today struck me a little funny, and I decided to investigate further.
Here are a couple excerpts from what Dave said about Verlander’s fastball:
One of the first things we noticed using that data this season was that Justin Verlander’s fastball disappeared in April. He was throwing 91-92 instead of his usual 94-95, and his performance suffered as a result.
For all the talk of guys learning how to pitch without their best stuff, Justin Verlander is clearly a better pitcher when he’s throwing 95 instead of 92.
It bugged me because I wasn’t sure it was true, either that Verlander’s fastball speed was improving as Dave said it was, or that there was a correlation between his fastball speed and performance.
So I decided to dig into the PITCHf/x data for Verlander. Here’s what we see about his pitch speeds going back to the 2006 playoffs.
During the 2006 playoffs his average fastball speed was 94.7 mph. In 2007, PITCHf/x recorded his average fastball speed at 95.0, although the period from which we have most of our PITCHf/x data is after the All-Star break.
In 2008, his average fastball speed has been 94.1, and the trend matches fairly well with that which Dave describes seeing in the BIS data. However, I’m not sure I see as direct a correlation between fastball speed and performance for Verlander as Dave Cameron does.
To look a little deeper, I calculated Verlander’s average fastball speed for each of his starts for which we have PITCHf/x data. I decided to use the Bill James pitching game score as the measure of performance, and I grabbed that data from Baseball Reference. (Fangraphs! Baseball-Reference! Is there any better time in history to be a baseball fan?) Comparing the game score for each of Verlander’s starts to his average fastball speed, there appears to be a correlation, but a fairly weak one. (The R squared is 0.09.)
I guess you could say he hasn’t pitched any great games with a fastball in the 92-93 mph range, although having a faster fastball does not appear to be a firm guarantee of success. Mostly at this point, I am skeptical of our ability to ferret strong conclusions out of a data set where the sources of error are on the same magnitude as the effects we are trying to measure. My skepticism applies healthily to the BIS data as well as the PITCHf/x data.