Most people use walks and strikeouts, the two most important aspects of pitching, to evaluate how good pitchers are, especially with prospects. While those two on their own do a much better job of projecting than ERA, groundballs are often overlooked. A great GB rate can offset mediocre BB and K rates; just ask Trevor Cahill and Lucas Harrell last season.
While location is a very important part of getting groundballs, average pitch height is not easily accessible, at least to us programming-illiterate. Using PITCHf/x, we can find the average vertical movement for a pitcher, along with an underrated factor, velocity. With just those two attributes, how much can we learn about a pitcher's GB rate? It turns out to be about 34% (R-squared), using starting pitchers with at least 60 IP last year.
While that is something, it's also enough to get you in trouble trying to use it for analysis. There are some big disparities between actual and expected GB rates, peaking at Andy Pettitte's 19.2% difference. It is interesting to see how pitchers with similar aggregate velocity and movement can have very different GB rates. Cahill and Chris Sale are a great example. They each averaged 86.5 MPH and had a VM value of 1.42, good for an expected GB rate of 51%. However, Cahill induced 61% groundballs, while Sale couldn't crack 45%.
Joe Kelly and Samuel Deduno had the highest expected GB rates, right around 55%, while Chris Young brought up the rear at 32.5%. In the components, Stephen Strasburg led in average velocity at 91.67 MPH with David Price as the only other starter above 91 MPH. Despite his "power" knuckler, R.A. Dickey had the lowest speed at 77.67 MPH, though Barry Zito, Justin Germano, and Jeff Francis all failed to exceed 80 MPH.
Germano had the most vertical depth, averaging -1.7 inches, while Deduno and Hisashi Iwakuma each stayed in the negatives. J.A. Happ had the most "rise," averaging +8.6 inches, while Young, Clayton Kershaw, and Jered Weaver also exceeded eight inches of rise. The most surprising part: Germano's actual GB rate was only 1.4% higher than Happ's last year.
It's easy to see the exceptions (aka limitations) of this model, but it would be interesting to see if year-to-year correlations are better using this type of model. It would be helpful to know that young pitchers like Kris Medlen and Scott Diamond may not keep up their 50%+ GB rates, due to their average sink and velocity. A 10% decrease in GB rate could mean a change in xFIP of around a-third of a run, worth about .7 WAR over a full season.