It's been a while since Part IV in this series, and since that time, I realized that I had been holding all of the information I needed to make a case that spin deception is a thing. The theory, most simply put, goes like this:
Wake effects deceive the batter by creating unanticipated movement.
The long-winded version is bit of a walk, but it's an easy one. Professional hitters have really good vision. Like, crazy good. They also have reps upon reps upon reps against live pitches. The result is that they are exceptionally adept at seeing and reacting to spin. Given the time hitters have to react to pitches at the professional level, it's practically a reflex. When a pitch moves differently than this reflex expects, the pitch is harder to hit.
Identifying Wake Effects
Since MLB switched over to the Hawk-Eye system, pitch tracking data has been elevated in a way that helps us identify this effect. The old "spin direction" measurement was really movement direction all along. With Hawk-Eye reporting directly on the pitch's actual spin direction, we can now fairly easily compare these two numbers.
Movement direction is no longer directly reported, so we have to dust off those old trigonometry functions and do some MLB-specific adjustments to get movement direction on the same scale as spin direction. Luckily, most coding languages and SQL have a handy atan2(y, x) function that does a lot of the heavy lifting. (I will answer emails and tweets about this math, but I won't further extend this lengthy post by elaborating on trigonometry.)
Once we have spin direction and movement direction, we can very easily figure out how far apart they are. The catch is that you can't assume a whole lot from such comparisons, and here's why.
A specific difference -- spin direction minus movement direction for a specific pair of values -- means something different for different tilts. An easy example is that a +30° difference for a pitch with a 180° spin direction (12:00 tilt) means it had a 150° movement direction (11:00 tilt), but for a pitch with a 0° spin direction (6:00 tilt), a +30° difference means it has a 330° movement direction (5:00 tilt). At first glance that's going to seem perfectly logical to you, but the first pitch moves to the right (batter's perspective) more than expected and the second pitch moves more to the left than expected!
A further wrinkle is found on sliders, tight curveballs, and any other pitch that finds itself in a gyro cluster near 0 vertical movement and 0 horizontal movement. Pitches with high gyroscopic spin are extremely sensitive to small variations in spin direction (tilt). This leads to exaggerated differences between the two direction angles. You set your sample data to exclude certain pitch classes, but that merely reduces the geometry problem instead of accounting for it. So...
Let's factor in movement distance. There are two "clear" approaches to this. You can be completely serious, like Glenn Healey and Lequan Wang, and use physics to reasonably calculate the side force, or you can be like me and use a somewhat reasonable alternative that assumes equal movement at both angles and measures the difference between the two endpoints. Healey and Wang definitely have the more accurate mathmatical approach, but while they are asking, "How much did side force affect this pitch?", I am asking a less specific question: How far did this pitch wind up from where the batter expected it to wind up? (The difference in approaches results in my values being smaller than Healey and Wang.)
I decided to express the directions as Spin Tilt and Movement Tilt since "tilt" is well understood and widely used. I settled on the name Tilt Difference for the difference between them and Deception Distance for the final value. But that's not all...
I did a truly excessive amount of thought before I touched any numbers, and decided that I also wanted to look at the absolute value of these differences. It didn't seem right to me that a pitch that generally has either +2" or -2" Deception Distance should average to a near-zero value because in reality there is an average 2" difference! This value gives us a second way to look for hidden value. We can check not only the average Deception Distance, but also the average Absolute Deception Distance.
I published a spreadsheet with all of this data: Spin Deception Data. It's a pretty neat little toy that lets you use the Data -> Filter views selection to shuffle between pitcher handedness and pitch type. The dataset is the complete pitch type summary set for 2020 as reported by the public MLB StatsApi. The filters are limited to pitchers that threw at least 20 of the filtered pitch type. Some dependent variables included in the sheet are swinging strike rate, foul rate, and in-play rate broken down by batter handedness as well as overall exit velocity and launch angle.
The second sheet in that document contains scatter plots with trend lines. Here are the most intriguing results, which I'm sure greatly please @NotRealCertain. The first set of charts is sinkers thrown by RHP.
Those are some pretty strong trendlines suggesting a positive relationship between both versions of Deception Distance and how poorly a ball is hit, and we see them again with the LHP sinkers. The fun thing about the LHP sinkers charts is that the trendlines are in the opposite direction because LHP sinkers have "positive" Tilt Differences.
I'm not an analyst, and I just do this for fun, so this is about as far as I want to take things myself. I'm sure there's plenty more to dig into, but it will have to be one of you that does it. If you have questions or comments, feel free to reach out to @texasleaguers on Twitter or use my contact form to send me an email.
Feel free to download, copy, and reuse the data in the spreadsheet, but please credit me, the blog, or the website if you publish any analysis related to the data contained therein.