Meet THE BAT X for pitchers… and THE BATcast Stuff Model!

I started this project nine and a half years ago. I visited multiple remote developers, overcame countless hurdles in an unknown number of hours, taught myself a lot about coding and modeling, and studied thousands of variables. But finally, six years after introducing THE BAT THE BAT is finally here!

Similar to THE BAT (Get it, because it’s THE BAT using Statcast data?)

Where can I find THE BAT X and THE BATcast stuff model?
Here at FanGraphs! THE BAT projections pagesthe player pagesthe play-off chancesand the auction calculator. THE BATcast is currently being added, so hopefully you’ll see it soon.

When you say ‘things’, what does that mean?
I mean the physical characteristics that describe a pitch. How it moves through space. Most people associate “things” with high speed, crazy breaking balls and lots of smells. While things can manifest that way, good stuff doesn’t necessarily mean a high strikeout rate. THE BATcast model is actually made up of numerous separate models, each predicting a different aspect of the game for a different type of field. I predict strikeouts based on things, yes, but I also predict walk rate, home runs, BABIP, etc. So if people are upset about that Framber Valdez is the top 10 on the BATcast list, they need to understand that we’re not just talking about swings and misses here. And the variables that describe things go beyond the obvious speed, motion, etc. that is plain for everyone to see.

What’s in THE BATcast and what makes it different from Stuff+/PitchingBot?
Part of the reason it took so long to build is because I didn’t want to just reinvent the wheel. The stuff systems that emerged in that 9.5 year period (like Stuff+ and PitchingBot) are great. If I was going to keep trying to build something, I knew it had to be unique in some way, otherwise what was the point? I could just use an existing system and integrate it into my projections.

As a starting point, THE BATcast contains all the variables you would expect from a stuff model: speed, movement, spin. It includes newer developments such as seam-shifted wake and pitch mirroring and approach angles. It looks at the interactions between a pitch and the pitcher’s primary fastball (not just how fast his slider is, but also the speed difference between the fastball and the slider). It implies that different pitches are better or worse against hitters of a given skill. However, all this is not new territory. Without giving too much away, I think THE BATcast does three important things (as far as I know) that others aren’t doing yet that I’d like to share.

First, these basic variables are used, but they are also put through a series of adjustments to try to squeeze more information out of them. For example, a pitcher’s slider with a side arm may have more movement than an over-the-top pitcher, but the hitter is likely expecting this based on the arm angle, so extra movement may not really help. THE BATcast adapts to account for unexpected movements, movements that go beyond what you would expect from a pitch given arm angle, speed, etc.

THE BATcast also contains tunnel variables. Tunneling is a fun word that has been used a lot in recent years, but I feel like it’s used as a catch-all term to mean a bunch of different things that aren’t really tunneling. I’ve heard it describes the release point difference or motion difference or other things like that. But when I use the word tunneling, I’m talking about looking at the entire trajectory of a given pitch and comparing it to the trajectory of the pitcher’s other offerings. Does the pitch hide within the path of the fastball, making it harder for the batter to pick it up? How much of the path is it hidden from, and how hidden is it? Does it stay within that path until a certain point and then deviate sharply after a hitter has already had to make a decision about the pitch? In other words, how similar are the pitchers’ pitches at different points or distances along the trajectory, and how does that mislead the batter?

The third thing I think sets THE BATcast apart from other systems is that it tries to take bridge heights into account. The most common bridge stitch is a cutter. It bridges the gap between a pitcher’s fastball and the slider, so that instead of being able to look for two different shapes, the hitter now has to look for three shapes that all flow into each other. Instead of a fastball running to the left and the slider running to the right, if there’s a cutter in the middle that can do a little bit of both, you’ve now created a wide range of possible motion for a pitch, making it much harder for the hitter to identify what’s coming. Even if the cutter itself isn’t good, simply having it can make the slider and fastball better. As with tunneling, I looked at the trajectories of different pitches in a pitcher’s repertoire, and a whole host of variables within them, to try to account for bridge heights.

That sounds like a lot of things…
It’s a lot of things (no pun intended). Because I’m looking at the entire range of the pitches, there end up being hundreds or thousands of tunneling and bridging variables that describe a single pitch. All told, the system evaluates roughly 4,000 variables and narrows them down to the most impactful.

Why does THE BATcast like Player X more than Stuff+/PitchingBot?
As I mentioned, the process of building THE BATcast starts with 4,000 variables, with each variable describing something about a certain pitch, before the modeling process narrows them down to the most important ones. But with the power of machine learning comes the frustration of its black-box nature. It can deliver accuracy and results, but it is much more difficult to put your finger on why it achieved a certain result. We can do our best to interpret, but unfortunately there are no clear answers to these types of questions.

Why does THE BAT X like Player Y more than THE BAT?
The simple answer is that he likes his stuff more than his superficial grades. Due to the black-box methodology, it may not be easy to identify exactly what it likes.

Are there any planned upgrades?
Yes! This is the system’s debut and there will be plenty of upgrades and refinements. Some of the bigger ones planned include height adjustments, reliever-to-starter conversions and creating a “stuff” system for minor leaguers. I’m also finalizing the seasonal version of this, which will hopefully allow us to identify fundamental changes in a pitcher very quickly without having to wait for the surface stats to stabilize.

…

Finally, I would be remiss without saying thank you. Dr. Alan Nathan, Tom Tango, Alex Chamberlain, Eli Ben-Porat, Eno Sarris, Nick Pollack, Mike Petriello, Thomas Nestico, Josh Kalk, Mike Fast, Harry Pavlidis, Dan Brooks, Jak Jones of Deuce of Diamond Analytics, and a host of others have all given up their time to explicitly answer my questions or simply provided inspiration for this project, and for each of you I am deeply grateful.