I've been looking at a way to get at arm angle and location on the rubber via PITCHf/x. I've added height to my database, just for the Cubs at the moment, to help put some stuff together.
When looking at release points, I want to just look at height of pitcher vs. release height (z0). For now, that is, as I'm trying to see if I can get some basic groupings that way.
To avoid noisy data, I'm only aggregating pitches that are fast enough to not have the high release on some curveballs. Yes, that is quick-and-dirty. A real run requires proper pitch IDs and corrected release points.
For this quick-and-dirty trial, with just Cubs, I only look where y0=50, start_speed > 80 and only at games played at Wrigley Field.
There is game-to-game noise there, but, to avoid the correction effort, I'm assuming the fluctuations are roughly the same across each player (not likely, given the varied sample sizes). Lazy.
For what it's worth, I think it shakes out pretty well as a first pass.
Here's the same thing, but the bubbles are now sized according to the distance from the middle of the rubber (x0)—righties usually have negative numbers (which is left from catcher's view, naturally), so those sizes are based on absolute values.
Using these charts as a "guide," here's what you can estimate—actually, just from the first chart:
The Guys Who Throw Overhand
Three-fourths, at Varying Degrees (going from more overhand to more sidearm)
A Lower Arm Slot, Lower Than Three-fourths
Things that can mess this up, not already mentioned, include pitching style (more or less aggressive leg drive) and posture, presumably.
Well, something is messed up, but I think it gets pretty close to the target. More information (e.g. x0 values) and more, cleaner, data are needed to make any real sense out of this.
If it starts to approximate reality as I tune and improve, I can try and incorporate it into pitch identification and plotting.
So, I hope you've enjoyed today's edition of Thinking Out Loud.