I’ve been trying to build a linear model for estimating slider quality, and it gave me some fantastic results… but there was one glaring issue that invalidated the entire formula: Kershaw’s slider graded out as one of the worst in baseball. That meant I was missing something really important, so I set about to search for the missing ingredient. It turns out that Kershaw does this ONE WEIRD TRICK, which may explain why his slider is so good.
Kershaw is exceptional at making the release of his slider look as much as possible like the release of the fastball. Let’s make that a little more precise, quantify it, by going step by step should you want to re-produce.
First, I filter to all SL pitch types with HB/F (horizontal break per 0.4 seconds) of 6 inches or less.
VRA = - ATAN (vz_r/vy_r) = Vertical Release Angle
HRA = -ATAN (vx_r/vy_r) = Horizontal Release Angle
Depending on the software you use to compute the above formulae, you’ll likely want to convert the number from Radians to Degrees.
If you download the savant data, you’ll see vx0, vy0 and vz0, from which you can derive the actually release velocities using physics. Let’s go through it step by step, though you can probably use the _0 (at 50 feet from the plate) variables as they’ll be really close to the _r (release point) variables.
tr = (-vy0 - sqrt(vyo² - 2 * ay * (50 - y_release)))/ay
This will give you the amount of time that elapsed between the actual release point and the 50 foot from the plate point. The release velocities can be computed:
vy_r = vy_o + ay * tr
vy_x = vx_o + ax * tr
vy_z = vz_o + az * tr
If you want to check that your math is correct (or that I didn’t mistype), you can hop over to Alex Chamberlain’s pitch leaderboard. Alternatively, if you get close to -2.1° VRA and -1.7° HRA for Clayton Kershaw’s slider in 2023, you’re probably gold.
The Key to A Deceptive Gyro Slider = Release Angle Synergy
Vertical Release Angle Synergy
Let’s define VRA Synergy as [VRA] - [FF VRA], where FF VRA is the average VRA of the fastball for that game. This gives us the following metrics:
The story is somewhat complicated since we have heavy correlations to location; the smaller number at the bottom is the average plate height. The sweet spot appears to be right around 0.4°, which we can see if we look at how well the pitch performed, while stripping out how well the pitch does given the type and location.
This isn’t a perfect measure, and it is noisy later on, but it does appear to confirm that 0.4° is the optimal number, which means if your slider has a -2.1° VRA, you want the fastball to be around -2.5°, which SURPRISE, that’s exactly what Kershaw has been doing for quite some time now.
Horizontal Synergy - RHP
The sweet spot here looks to be around 0°, which SURPRISE, is exactly what Kershaw has been doing. Here’s the same chart, but for LHP:
Deception / Release Angle Synergy: Formula
SQRT(
SQUARE([VRA]-[FF VRA]-0.4)
+ SQUARE([HRA]-[FF HRA])
)
The closer to 0, the better. Here’s where it gets really interesting:
The small number at the bottom is the average height at plate crossing. Despite the numbers being close to the bottom of the zone, we see more swings! This is such a strong signal for deception that frankly I’m worried I messed something up. For my sanity, I would love it if someone could independently come up with a similar result.
Unsurprisingly, pitches that optimize for Release Angle Synergy outperform similar pitches, when we strip out location.
Here are 2023 SP leaders for Gyro Slider Deception:
And the worst SPs in 2023 by this metric:
Too Good to Be True?
As I mentioned above, I’m worried it’s too good to be true, but I’m hopeful that it’s real (and spectacular). If I messed up, I will be eternally grateful to you for correcting the public record. If I didn’t and you can re-produce, I will be immensely grateful as well. More to come on the subject of Gyro Sliders (and RA Synergy), but I felt this was important enough of a piece to warrant publishing right away.
I think this is very interesting study. Thank you for publishing it.
I have one question. I think VRA and HRA vary depending on the course and height of the pitch. How are the VRA and HRA of each pitch calculated?