PITCHf/x is an incredibly useful tool that helps fans, analysts, front office personnel, and anyone else who is interested understand exactly what and where a pitcher is throwing. Much of my analysis leverages BrooksBaseball.net, an unparalleled resource when it comes to interpreting PITCHf/x data and creating outputs that are useful to both analyst and casual fan alike. Brooks Baseball is curated by Dan Brooks and Harry Pavildis (among others who will be mentioned later on).
Discuss this post & PITCHf/x on the BSL forum here.
For example, to learn more about O’s starter Scott Feldman, who started for the O’s in Arizona tonight – one could simply check out his Player Card. You’ll see at a glance what pitches he throws and the rough velocity for those offerings. You can also dive much deeper into the data by clicking on any of the callouts below, depending on what you’re looking for.
I’ve put together some rough points to understand when learning about PITCHf/x, but reading any pieces linked to will provide a much more in depth description of the service.
Mike Fast‘s article for the Hardball Times Annual outlines what exactly PITCHf/x is and how the data is compiled. As you can see from his twitter account, Mike has been hired as an analyst by the Houston Astros, so he clearly knows what he’s talking about when it comes to this sort of thing.
As Mike points out, PITCHf/x is a combination of some very objective data such as velocity, spin, location, etc. and some more subjective data such as where the strikezone sits for a given hitter. For more information on PITCHf/x and its components check out this helpful guide from Alan Nathan. It also outputs some subjective data as the pitch classifications are driven by an algorithm that can struggle with identifying the correct pitch classification.
Brooks Baseball takes classification to a new level, as they manually check every classification to make sure that they are accurate. This is largely Harry Pavildis‘ territory, and he’s got help from Dan Rozenson, another brilliant PITCHf/x mind. These guys take the raw data from MLB Advanced Media, pretty pictures of various pitchers, and scatter plots and determine exactly what pitches are being thrown and when.
That brings me to tonight, where a comment from the O’s broadcast booth lead to a long image search, and ultimately a better understanding of a pitcher’s repertoire and PITCHf/x in general. Gary Thorne and Jim Palmer mentioned in a particular at-bat that Scott Feldman throws a split-finger fastball “less than 1% of the time”. This seemed, well, odd to me. My first inclination was to jump on to Brooks Baseball to see what it had to say. I found nothing of this mysterious splitter. However, an exhaustive image search lead to this tweet, which was the beginning of a trip down a PITCHf/x rabbit hole.
August 31, 2012 – Source: Jason Miller/Getty Images North America
In this image of Feldman, from 9/1/2012, you can clearly see the split-finger grip he’s using to deliver the pitch.
Dan was next with a big discovery, as he found a scatter plot from a 2012 start where two distinct clusters are apparent. As Dan would point out, a split-finger fastball typically has less fade than a changeup. As such, you should see two distinct clusters with the difference being horizontal movement. Previously many of these pitches that we now know to be splitters were classified as changeups. Their velocity is the same as Feldman’s change, and typically guys that throw changeups don’t also throw splitters (and vice versa as Dan would note.
I took a minute to clarify what Dan was looking at, so that it’s a little more clear what exactly he meant by “two clusters”. As you can see in the link in that tweet, there are two distinct clusters with the “split” cluster having a few inches less horzintal movement than the changeup cluster.
This process helps improve what we know about Feldman’s offerings and how PITCHf/x classifies pitches. With every instance of something like this our understanding improves which is ultimately the goal. There are simply too many pitches, and too many nuances of pitcher grips and pitch types to be 100% accurate all the time. This crowdsourcing helps improve the data for everyone and that’s how you can help.
Now that you have (hopefully anyway) an understanding of PITCHf/x and how the various pitches are classified, you too can help identify discrepancies or nuances that can improve our understanding of this aspect of the game.
Jeff was the owner of the Orioles blog Warehouse Worthy, which focused on making advanced statistics a part of the conversation for the average fan. Outside of baseball, Jeff is a graduate of Loyola University where he received his Bachelor’s and Master’s in Business Administration. The Maryland native currently works for an Advertising Agency in downtown Baltimore. Previously a contributor to Beyond the Boxscore, he joined Baseball Prospectus in September 2014. You can reach him at jeff.long@baltimoresportsandlife.com.