Swing plane has always been an important metric for evaluating batters, but there has never been an objective measure of swing plane in the public. Privately, through wearable technologies or video analysis one could obtain information regarding swing path, which has since gone on to become a valuable tool for coaching and training. Even still, we lacked data from actual games.
Over the past two years many writers and researchers have turned to using Exit Velocity vs Vertical Angle charts to analyze batter performance. Ranging from Rob Arthur to David Kagan to Alan Nathan. It is a great way to visualize the data. I have wondered in the past whether you could use these information to estimate the average plane of the bat on impact. If you were to plot a second order polynomial regression on top of such a chart and take the derivative of the function you could find the peak launch angle, that is the angle with the highest average exit velocity. Perhaps. I’m not entirely sure this is the case, but it is a good place to start.
As you can see in the chart below, Freeman has had roughly similar “peak” angles and “peak” exit velocities over the past three seasons. However, it might appear that 2017 is a tad lower than the other two. Which should make some sense, considering he broke his wrist part way through the season.
As you can see in the chart below, Freeman has had roughly similar “peak” angles and “peak” exit velocities over the past three seasons. However, it might appear that 2017 is a tad lower than the other two.
I have gone through and found second order polynomial regressions for every batter in each of the three seasons on record. I have found their estimated peak angles exit velocities, and I put them into a viz that I hope you will enjoy.
One thing I want to make clear: this exit velocity is not observed velocity, it is the exit velocity produced by the regression model. I hope that this exit velocity number is a function of average pitch plane (the angle of the pitch with regards to the horizon), plate discipline (how often the batter swings at pitches that are far away from his bat plane, ie difficult to hit), and bat speed. My hope is that this velocity is proportional to the bat speed, that is my true goal here. Currently, I have no way to test this theory, but hopefully I will be able to in the future. For now, let’s just roll with my assumptions.
If you click through the various seasons, you can see how the peak angles climb in each of the past three seasons. The 2015 season clearly has the lowest angles, the 2016 season is spread throughout the middle, and the 2017 season clearly has higher angles. On average. Not every batter follows this rule, and the change is reasonably subtle, but generally speaking this is what we have seen across major league baseball.
(click the gif to make it larger)
Anyways, you can play with that viz, but I can also show you a table for guys who have undergone the largest bumps in peak exit velocity over the past three seasons.
A few of these guys may not be the most fantasy relevant guys out there. However, there are a few good names. Wilmer Flores may be a bit of a shock to some. His bat is solidly underrated, but he has killed left handed pitching lately, and the Mets currently have holes all around the infield which could give Wilmer a lot of potential plate appearances this season.
Ketel Marte popped up on Jeff Sullivan’s radar as a young breakout candidate. Jose Ramirez and Lindor both had monstrous seasons in 2017, likely fueled by this peak exit velocity bump. Guys who didn’t quite make the cut are Schoop (+2.3), Alonso (+2.1), Andrus (+1.3), and Ryan Zimmerman (+1.1).
I could go through the decreasers and angle changers, but instead I will simply link to a document containing all of the data. You can feel free to look through it at your leisure. I want to quickly cover one possible flaw in this analysis:
I spent a few hours this past weekend playing around with k means clustering these batted balls, and I found that somewhere around K=12 or K=13 was just about right, in terms of withins and betweens analysis. That isn’t important, though. I want you to look at where clusters may be sitting in this space.
This is a plot for K=13. Notice how there are six clusters at the top exit velocity regions around the perimeter. Three are ground balls (purple, red, light blue), one line drive (dark blue), one fly ball (orange), and one pop up (mustard yellow). At the far flanks you have extreme ground balls (camo green) and extreme pop ups (dark green). Those all make intuitive sense.
Next, you have five “inner” groups. When you change K from 13 to 12, this will drop to 4 inner groups. And I believe it drops to 3 if you lower to 11. Anywho, of these inner groups, you have three ground balls (magenta, teal, green), one pop up (pink), and one line drive (purple).
PS, I am very sorry if you’re color blind.
Of these five inner groups, that purple line drive group (the top right of the middle five) is exceptionally valuable. The other four are worth next to nothing. Furthermore, the bottom three groups are exceptionally rare.
These “inner five” clusters, especially the bottom three, could, perhaps, throw off the analysis of these Exit Velocity versus Launch Angle charts. Perhaps, I’m not entirely sure. This is a topic I am actively trying to understand better myself.
This weekend I uploaded a few videos to youtube that show 3D representations of this chart. You might find them interesting. This one shows the clustering in 3D. Below I have embedded a video that displays batted ball frequency with respect to exit velocity and launch angle. These are not high production quality videos, I recorded them on my desktop using shadowplay. But maybe you’ll find them interesting. I find it difficult to talk about 3D structures using 2D projections and words alone. The video might help.