Anatomy of a pitch: Change upby Josh Kalk
July 01, 2008
|Cole Hamels has a nasty change up but would it be more effective if he threw it a little faster? Phillies at Nationals, May 20th, 2008 (Icon/SMI)|
Last week while writing about Zack Greinke I noted that he produced a tremendous differential in speed between his fastball and his change up. John Walsh added that that might not necessarily be a good thing as he mentioned that if the speed difference was too great it might actually help a hitter detect the change up. I had assumed that the bigger the speed differential the better it was for the pitcher like a faster fastball or more vertical drop on a curve. But because a change up is supposed to fool the hitter into thinking fastball at some point too slow of a change up is going to be a problem. For instance, if a pitcher threw a 90 MPH fastball and followed that up with a 50 MPH change up you can bet the hitter is going to notice. So the question is at what point does this effect occur? If it is at a large differential like 20 MPH then most pitchers don't have to worry but if major league hitters are tipped off by a 8 or 10 MPH differential then this is a real concern.
Like previous pitch articles I would like to use simple Pearson correlation but if the data isn't linear these correlation coefficients aren't a good measure. So the first thing we need to do is to plot pitcher's speed differential against our old friend runs100 to determine if the correlation is linear or if hitters are identifying a large speed differential causing a nonlinear distribution. Here I am using pitchers who threw at least 100 fastballs and 100 change ups during 2008 though either sinkers or four seam fastballs are allowed. I am defining change up speed differential and the difference in speeds between a pitchers average fastball and change up. For graphical purposes I am actually plotting negative runs100 as a negative runs100 is good for pitchers but not as easy on the eyes.
The speed differential goes from about 5 MPH and guys like Ben Sheets and Aaron Laffey to nearly 15 MPH and guys like Scott Kazmir with Johan Santana and Cole Hamels around 12 MPH. This curve definitely is linear so it appears that at least out to about 15 MPH the bigger the speed differential the better. You can see a slight tend upwards and now that we have assured ourselves that the distribution is linear we can use the simple Pearson correlation coefficient. When we do that we get a correlation of 0.23 which is still considered a small correlation but relatively large when you go back and see what the correlations were for curveballs and sliders.
Correlating some of our favorite variables we find that release point finally seems to make a difference checking in at a correlation of 0.24. This is much higher than it was for curveballs and sliders implying that hiding the change up with a constant release point is more important that it is for other off speed pitches. This could be due to the fact that change ups move like a fastball so hitters will have a harder time trying to pick out the change ups from the fastballs once the ball is in the air. Interestingly, when we break down release point into horizontal and vertical components the horizontal part appears to matter much less than the vertical part (0.5 to .17 correlated). Are hitters looking for a different vertical arm position to tip off the change up? It appears possible.
When we correlate the difference in movement between the fastball and change up we see a slight advantage towards the pitch moving more horizontally and dropping than a true straight change that has little difference in movement from the fastball. This is because change ups are usually thrown to an opposite handed batter staying low and away. The extra movement here just magnifies things. Here, more horizontal movement is slightly preferred correlating at 0.13 and vertical drop correlating at 0.11 with total movement correlating at 0.14. You may wonder why the total movement correlates so poorly compared to the horizontal and vertical components but that is because total movement is just how different the movement is from the fastball so a change up that moves less horizontally and rises slightly more than a fastball it can have the same total movement as a change up that points in the preferred direction. If you apply a directionality preference then the correlation improves to .22.
Change ups are a staple off speed pitch especially to a left handed pitcher because they are so effective against opposite handed batters. Within reason, the more speed differential on a pitcher's change up the better. Hiding the release point and slightly tailing the ball down and out to an opposite handed batter are also important. Unlike curveballs and sliders where one or two things mattered a lot and the other variables hardly correlated, all of these things are about equally important for change ups. This likely adds larger demands on a pitcher who has to master the pitch before getting rewarded. You often hear that a young pitcher doesn't have a good feel for his change up and that likely means that one of these three things is out of whack.