Every other week I write an email discussing what I learn growing Reboot Motion. My Reboot Breadcrumbs posts are a look behind the curtain at some of the new things our team looks to deliver to our partners. If you want to receive these directly in your inbox, subscribe below.
Velocity trumps all.
Whether you are an athlete, a coach, or a biomechanist, we all believe the best way to improve pitching performance is to throw harder.
And Reboot is no exception. In fact, we’ve made this exact pitch:
While this slide is great for companies like us and is widely important in driving player development, I am not sure it makes a huge difference in player evaluation.
That is not because velocity doesn’t matter, but rather because everyone already knows that it does.
Contrarian is Key
Being right is important. But it does not help much if everyone agrees. This is true in investing- being a bitcoin believer the last 10 years was likely way more lucrative than the next ten will be- and it’s true in sports- signing great three point shooters was a lot more advantageous before the entire NBA put a premium on the skill.
This is why Moneyball worked so well 20 years ago, but does little to add value today. Advanced metrics are just as accurate as they were back then. In fact, they are more accurate. The problem is they are now consensus.
The Performance Checklist
So what should teams look for when searching for a hidden gem? How should scouts analyze various prospects on the road? And how should R&D teams comb through their data?
If the goal is to search for things that are 1) non-consensus and 2) right, let’s start with right. (If we can only get one, that is the more important one after all.)
Velocity. Despite the lead in, velocity trumps all. If I am in charge of player development, this is still where I would spend a large portion of my time.
But the fact that everyone knows it is true means that is not where decision makers will find their edge. It is a must have- but it is firmly in the “consensus and right” camp.Spine Rate/ Pitch Break. Similar to velocity, these metrics are 1) easy to measure and 2) generally seen as predictors of performance. It will be tough to find an edge here, but, for the same reasons as velocity, must be accounted for.
Deception. As Ben Lindbergh writes in his profile of Yusmeiro Petit titled Yusmeiro Petit and the Well-Hidden Power of Pitcher Deception, deception is “generally reserved for any tactic or trait that makes a pitch harder to hit than its flight through the air would indicate.”
In other words, it’s the stuff we can’t measure. It’s the variable decision makers hope to solve for when predicting future performance. Organizations that are early to decoding deception will fall into the non consensus and right bucket- where all the gains are made.
Deception
In Lindbergh’s article on deception, he touches on the endless list this broad term encompasses:
Tunneling- aligning pitch trajectories so pitch types look similar until it’s too late for hitters to react
Slot changing- variation in arm angles, which prevent hitters from predicting the pitch based on arm slot (basically the opposite of tunneling)
Ball hiding- how much an athlete can or cannot see the ball while it is in the pitcher’s hands
Long strides- that reduce the time from ball release to contact
Any rare trait that hitters are not used to seeing
And basically everything Johnny Cueto does
Eno Sarris describes just how “catch-all” deception is on the Driveline R&D podcast.
In short, there will always be things we can’t measure. Maybe there should be a term for that. Maybe there should not be.
But, if there is, deception may be the wrong one.
That is because if deception is the difference between what a batter predicts the ball will do, and what it actually does…and we believe it can be measured.
Measuring Deception
If hitters waited until the ball left the pitcher’s hand to start analyzing a pitch’s location, speed, and spin, making contact may be impossible. There is just not enough time when competing against the world’s best.
Rather, great hitters rely on the pitcher’s movement to pick up cues that expand the 400 milliseconds they’d otherwise have to make a decision.
In other words, hitters know what the ball should do based on a pitcher’s delivery. When their prediction matches what the ball actually does, advantage hitter. When it doesn’t, advantage pitcher.
—
Remember, this is all just one big physics problem.
In Reboot Motion’s current analyses, we compute each body part’s rotation plane- which is the primary plane in which it is rotating. Due to the conservation of momentum, the rotation planes of the body are the main input into the rotation plane of the ball- also known as the ball’s spin direction.
Therefore, the spin direction of a fastball can be pretty well predicted from the rotation plane of the pitcher’s body.
And this “implied” spine direction is what we think hitters expect to see.
That being said, this prediction is far from perfect. For example, some pitchers use their wrist, fingers, and other body parts (intentionally or unintentionally) to offset the rotation plane of the ball relative to the rotation planes of the body.
Our hypothesis is that hitters can observe the way a pitcher’s body is rotating, and somewhat predict the way the ball will spin. If the ball’s actual spin direction is different, we’d expect the pitcher to outperform relative to standard predictors of success like velocity and spin rate.
One of the more public examples is Josh Hader, who has a more side arm delivery, which implies a more horizontal fastball spin direction. However, due to how he uses his wrist and fingers (among other things), his fastball actually has a more vertical spin direction, giving it some rise.
Travis Sawchik dives deep into Hader and Baseball’s Most Mysterious Pitch, showing the difference between what batters expect and what they get.
To explore this concept further, Reboot created a fastball deception metric based on the difference between the spin direction of a fastball implied by a pitcher's delivery, and the actual spin direction of the fastball. We then aligned this metric with Statcast’s percentile rankings.
From here, we tested how well fastball velocity, spin, and Reboot’s internal deception metric predicted pitcher outcomes. Fastball velocity and spin were chosen to pair with our metric due to 1) the common belief they are good predictors of success (and the fundamental physics to back it up) and 2) their availability in the aforementioned data set.
Using a K Nearest Neighbors Regression approach, we found that these three metrics could predict whiff percent- a term that measures a pitcher’s ability to create swings and misses- with an r^2 value of 0.38.
This isn’t as strong as it could be because many other factors should go into a proper model of whiff percent, but we felt this was a solid result given our limited data set.
(Remember, in our “Breadcrumb” posts we are discussing our R&D work as we do it…there is much more work ahead of us.)
Below is a graph of the importance of each feature in the model.
While our deception metric is a clear third to fastball velocity and spin, we expected this. The important thing is that it matters- contributing significantly to the predictive power of this model.
The Invisiball
Circling back to Lindberg’s profile on Yusmeiro Petit, he showcases Petit as a textbook case of someone that has “it”, whatever “it” is.
Petit clearly does something different than most MLB pitchers- he wouldn’t have had over a decade of success against the world’s best while throwing under 90 mph otherwise.
Lindbergh credits Petit’s “invisiball”, described by Angels TV analyst Mark Gubicza below:
“He hides the baseball so well,” Gubicza said. “You’re at the plate and you see a pitcher hide the baseball, keeping that front shoulder in, and you see the baseball out of the hand and you’re thinking, ‘OK, it looks just like a fastball.’ Changeup! … If you have the baseball out in your hand, if you’re not as deceptive as Petit is, you can at least see the grip. … But when you’re hiding the baseball and all of a sudden as a hitter you’re reacting to seeing the ball, it’s already out of the hand by the time you get a chance to recognize between a changeup and fastball.”
Hiding the ball may help. It clearly works for Petit.
However, I am a little skeptical that such major outperformance is simply the result of playing hide and seek with the baseball.
With this in mind, we wanted to see if something else was going on. Specifically, we wanted to see where Petit scored based on our internal metric. And, with the power of public video and our own motion capture technology, we ran the numbers…
We were not surprised to see him score around the 90th percentile.
The Takeaway
So has Reboot Motion solved deception? Unfortunately, no.
We have more work to do. We expect by looking at additional body metrics and various performance benchmarks we will improve our deception model over time.
(We also expect our partner teams’ R&D departments- who are already measuring what matters- are working on this, and keeping their discoveries to themselves.)
But today, we can be confident in the following:
There is a relationship between body rotation plane tilt and ball spin, which allows us to calculate a “predicted spin”.
This predicted spin is similar to what batters are implicitly “calculating” prior to ball release.
Pitchers that score high on the internal Reboot deception metric tend to outperform.
Even if deception is not as strong of a predictor as velocity or spin rate, it still matters. And you have to measure what matters.
Remember, in any industry- especially the most competitive ones- the spoils go to those who are 1) non-consensus and 2) right.
Nowhere is that more true than baseball.
It happened the last twenty years with Moneyball, and it will happen the next twenty with biomechanics.
We expect the best teams to be those that can predict success better than the competition…and that starts with quantifying what was previously unquantifiable.
Almost all sustained success in competitive industries is based on one's ability to convert information asymmetries into real world value.
I call this information arbitrage.
Applied biomechanics is this for ALL athletic competitions, but different from one sport to another ... and different between positions in a sport.
We may find there's a particular correlation between being a world class diver and a championship wrestler ... or point guard and slot receiver ... thereby allowing athletes to identify complimentary training that doesn't overly stress their main mechanical movements and/or allow athletes who are not quite 'there' in one sport to transition to another and, in so doing, continue to compete at a high level.
Nice article!