This article will cover our pitch design process for the fall of 2019. In another post, I will go over different goals for each pitcher (anonymously) and their most updated profiles.
Having team practice prior to individuals gave almost every pitcher (unless injured or shut down for games) the opportunity to showcase their stuff in competition prior to their bullpens. In a couple cases, this was very useful because their in-game stuff played much better than we would have expected just by evaluating their Rapsodo data.
One of the first things we did this fall was get video/pictures of each pitcher’s pitch grips to see the “before” if a change was made. We also wanted to get the best slow motion video we could, granted it was with an iPhone.
The first thing we did with each movement profile was create an easy visual for the players to understand and for our pitching coach to evaluate. The visual came in the form of overlaying their Rapsodo movement profile with a clock as seen below.
After we overlayed the two, we drew a line through the middle of the fastball cluster to show the average axis extending out to the clock. In this case, the average fastball axis would be between 1:00 and 1:15. Extending that line through the center showed us what axis a breaking ball that “mirrors” the fastball would be. Think of mirroring like this: two different pitches having the exact opposite spin, such as a fastball with an axis of 12:00 and curveball with an axis of 6:00. This particular pitcher happens to have a curveball that mirrors the fastball, and the line through the cluster shows the axis between 7:00 and 7:15.
After visualizing whether or not the pitches mirrored and if they have had previous success with that pitch, we evaluated the separation between pitches. Separation is what it sounds like, having movement separation/differential between other pitch types. Below are two examples of separation. The first profile lacks separation between fastball/changeup/sinker, as you can pretty much only see a brown cluster, and the second profile has good separation between fastball/changeup. Shifting the axis of the sinker/changeup to be closer to 3:00 would be the goal for the first profile if he has not had success with it, in order to create more separation.
An important note about our process is that we do not use a cookie-cutter approach. We would never have all pitchers stop throwing their 2-seam fastball and start throwing 4-seam fastballs, which believe it or not, has been a thing. To do that makes minimal sense. You would essentially be saying guys that are in the MLB Hall of Fame for throwing a 2-seam fastball would have been better off throwing a 4-seam instead. And then there's the psychology side to that. What if a pitcher believes that his 2-seam is his best pitch and all of a sudden you strip that away because data says 4-seam fastballs are performing better? There are outliers and exceptions, and I believe that is one of the shortcomings of modern analytics.
One of the biggest things to understand with pitch design is that there is not necessarily a “right” answer and there is many exceptions to what was covered above. It would also be naïve to tell a pitcher to adjust their grip, having likely thrown it a certain way for years, without explaining to them the potential benefits. Presenting the clock as a visual (like the picture above) is a simple way to show some potentially beneficial adjustments, and a way we can show them their changes over time.
How We Can Improve Pitch Design
Moving forward, there is a lot of improvements we can make. “Mirroring” and creating separation are great concepts, but meaningless unless they can be executed in games. I now believe that "tunneling" is drastically more important than mirroring. We can see if pitches tunnel a few ways--if they mirror, there is a possibility they tunnel but we are understanding more and more than velocity plays a major role in that. The easiest way to see if they tunnel is by overlaying video both from behind the pitcher and catcher. The more complex way is by creating the plots below. Unfortunately, Trackman data is needed for this and we only have a few pitchers with that data. Andrew Smith, @roo1776, wrote the code to make this possible. The greenish-yellow circle indicates the hitter's decision point. Two perfectly tunneled pitches would have the same movement up to the decision point, and separate to the end location (red circle) after the decision point. *Note for below: The blue line is a 2-seam fastball, the orange line is a slider.
It's also important to point out that tunneling has a lot to do with locating pitches. For example, in order to tunnel anything with a true 12-6 curveball, you will need to be able to throw your fastball up in the zone to give the illusion of a high fastball when throwing the curveball. A fastball located at the bottom of the zone will not tunnel with a 12-6 curveball, but that does not make it a bad pitch.
Rather than basing everything off the fastball, we should base everything off the best pitch. For example, a pitcher whose best pitch is their sweeping slider should not be working to increase the vertical break on their fastball. By shifting the axis closer to 12:00 (which is a common goal), the fastball would allude cut due to the high quality slider compared to a fastball with more horizontal break to pair with a slider.
Jacob Foster has written various articles about how these traditional ideas of pitch design have their flaws, and demonstrates how in general, secondary pitches with more horizontal movement perform better than those with vertical movement.
We also had some cases where movement profiles made natural changes as pitcher mechanics changed and they began to move differently. For example, one freshman pitcher became more glute-dominant as the offseason progressed which changed his throwing posture and shifted his fastball seam orientation closer to 12:00. Pitch design "answers" are not always within grip adjustments or pairing it with another pitch.
In a later post, I will detail how we scored each pitchers' progression and the relationship between their scores and the type of learner they are. The goal there will be to try and select the players who will not only benefit most from pitch design, but how their goals are communicated.
Comments