Data Driven Deception
Every other week I write an email discussing what I learn launching and growing Reboot Motion. If you would like to receive it directly in your inbox, subscribe below.
A GM has one job- predict the future.
Whether through the MLB draft, re-signing current players, or staffing the organization with great coaches, scouts, biomechanists, and more….they are looking to put together the group that gives them the best chance to win.
And given the size of MLB organizations, and more importantly the scope of potential targets, they always look to turn qualitative opinions into quantifiable data.
Stuff Plus
One area where baseball has made tremendous strides is pitching analysis. While we all can see who the great MLB pitchers are, predicting future production given current performance at various levels is not as easy.
Analyzing college pitchers, international prospects, and high school talent…and normalizing what they see to to predict the future is no small feat.
To help solve this problem, front offices turned to Stuff+, which Driveline Baseball describes below:
The art of “pitching” has many variables associated with it, some easily quantifiable and others not so much. Every pitcher, coach, and organization is constantly in a battle of weighting these variables in a manner that maximizes development and performance on the field.
While there is little agreement on how important “secondary” variables such as deception and sequencing are towards generating outs, it is generally accepted that three pitcher talents stand out above the rest.
The ability to:
Command the baseball.
Generate ball velocity.
Manipulate the baseball and acquire unique movement on pitches.
While people have different models, Driveline and others mainly focus on a pitcher’s:
Pitch Velocity
Vertical Break
Horizontal Break
Arm Angle
Release Extension
The belief is, by analyzing these characterizes, we can strip out variance, the level of competition, and other less predictive variables to predict the future.
The Outliers
While we generally see strong correlations between Stuff+ scores and outcomes variables like ERA and WHIP, every once in a while there are outliers- pitchers who continue to excel despite a lack of velocity, a lack of movement, or simply a lack of “stuff”.
In 2021, Ben Lindbergh profiled one of these outliers- Yusmeiro Petit for the Ringer, asking the simple question “why is this guy good?”.
The answer came down to a far less simple, but all encompassing answer: deception. If Petit isn’t beating batters with velocity, and he isn’t beating them with movement, he must be doing something else. He must be deceiving them.
This past November, Michael Ajeto of Baseball Prospectus was on the Effectively Wild podcast, detailing Cristian Javier. Javier, like Petit, was continuously outperforming expectations from traditional Stuff Plus models.
The question was again a simple one: how?
Defining Deception
Ajeto hypothesized Javier’s success is at least somewhat due to a difference in 1) the ball’s vertical approach angle and 2) his arm slot. Jeff Passan agrees, calling it a unicorn pitch.
Their explanation is as follows: MLB pitchers throw fast. Really fast. In order for hitters to compete, they pick up cues from the pitcher’s delivery prior to release. If the ball behaves differently than hitters expect based on the pitcher’s movement, they will have a harder time making contact.
They will have been deceived.
At Reboot, we’ve looked into defining deception in the past. But now, due to our integration with TruMedia, we can do even more.
Data Driven Deception
With the help of TruMedia, we wanted to dive into Ajeto and Passan’s claims and compare 1) Javier’s delivery with 2) the ball that comes out of it.
We did this by first looking at the Reboot metrics that most closely resemble arm slot. Javier’s fastball arm slot is as close to 3/4 as one can get, centered around 1:30 on the clock below.
However, the ball’s approach angle is much more vertical, centered around 12:45- equating to an approximate 22 degree difference.
Next, we created a singular metric to measure the difference between arm slot (or what we call “pitch hand hilt”) and vertical approach angle, labeling it “Pitch Hand Ball Angle Diff”.
In the graph above, pitchers at the very top or very bottom have the most extreme differences between arm slot and vertical approach angle.
Those at the top have a less vertical approach angle than expected, while those at the bottom deceive batters with a more vertical approach angle.
It should be no surprise to see Javier at the very bottom.
Stuff Plus Plus?
So what does this mean?
It doesn’t mean we’ve “solved deception”, but it does mean there may be another quantifiable metric to analyze performance.
We (and everyone else) still think velocity and movement are the two biggest drivers of success.
But, as Lindbergh puts it in his profile of Yusmeiro Petit:
Pitchers whose deception makes their middling movement, speed, and spin play up exhibit all the ingredients of a good old-fashioned market inefficiency.
And we believe the first step in capitalizing on a market efficiency is quantifying it.