Wednesday, January 20, 2021
Divergences
Jacob deGrom throws his changeup with slightly more sidespin than backspin. You can see it in the image on the left (batter vantage point), where the changeup is released at 10:15. However, based on how he throws the ball, by the time the ball reaches the plate, its average path will have the ball moving almost totally to the side. As a result, the batter will be observing a pitch that will move at 9:15 (almost all tail). That’s a pretty large amount of divergence or deviation. You can see it right there, a one hour deviation.(*)
(*) For those who prefer degrees: there’s 12 hours in a clock and 360 degrees in a circle. So each hour is 30 degrees. A one hour deviation is a 30 degree deviation. Use whichever you are most comfortable with.
Now, Jacob deGrom is not perfect. He does not throw his changeup the same way each time. Sometimes he’s able to convert all that spin to the side. But sometimes not. Does it matter? Well, probably we’d guess. Ok, the job of a saberist is not to see if some effect exists. We’re dealing with humans: some effect will exist (even elusive traits like clutch and chemistry). What we’re really after is the MAGNITUDE of an effect: HOW MUCH does it exist?
So, I took the 161 deGrom changeups that have the necessary tracking, and I split them into three groups: the 30% with the least amount of divergence or deviation (group 1), 30% with the most (group 3), and the middle 40% (group 2). I have run values for each pitch.(**) Negative run values lowers run scoring and so is good for the pitcher. Positive run values is bad for the pitcher.
(**) How to explain run values… let’s see. Every pitch will either advance the count, or end the plate appearance (walk, strikeout, hit batter, hit into play). When a pitch advances a count, there’s a small benefit to the batter or pitcher, depending on ball or strike. When it ends a plate appearance, there’s a big benefit to one or the other. Each combination of ball, strike, runners on base, and outs has its own run expectancy, its own state. After each pitch, we have a new state. And the change in those states has a change in run impact. That change is the run values.
For deGrom, his changeup in group 1, the group that has the least amount of deviation, has the run value increase by a whopping 1.98 runs per 100 pitches. If deGrom would normally give up 2 runs per 100 pitches, then throwing his changeup with limited deviation would make it 4 runs per 100 pitches. His changeup in group 3, the group with the most amount of deviation, the run value is a minus 6.55 per 100 pitches. This is an insanely great, (literally) off the charts great changeup. The group 2 changeup is a modest -0.16 runs per 100 pitches.
We see a good effect on his slider as well. He “converts” a ton of the initial backspin of his slider so that it moves to the side. His group 1 slider is -0.38 runs per 100 pitches. But his group 3 is -2.93 runs per 100 pitches. Even his group 2 is -2.63 runs per 100 pitches. So for deGrom, he doesn’t even need to have that much deviation, just some. When it’s flat, it’s still an ok pitch, but not devastating.
As for his 4-seamer: it’s his “average” 4-seamer that is the most effective. His group 2 is -3.80 runs per 100 pitches. His group 1, the most flat, is -1.04, while his group 3, the one with the most deviation, it comes in at -0.65 runs.
All pitch types and pitchers
Since we did it for deGrom, we may as well do it for all pitchers and all pitch types. I took each pitcher’s pitch type with at least 150 pitches. And I broke it down into little to most deviation. We’ll start with the least interesting one, the 4-seam fastball. Of the 217 pitchers in my sample, those in group 1 (least deviation) is -0.17 runs per 100 pitches. Group 2 is -0.11 and Group 3 is -0.14 runs per 100 pitches. In other words: there’s no general statement we can make with 4-seamers as to whether alot of deviation is good or bad, desirable or not. For some pitchers, like Aaron Nola (+2.91, -2.38, -4.37, respectively for groups 1, 2,3) you want alot. For others like deGrom, you don’t want anything extreme.
Sliders are similarly boring, deGrom notwithstanding: -1.12, -0.94, -1.12, respectively for groups 1,2,3. Again this is league wide and what might help one pitcher in extra deviation would hurt another pitcher.
Curveballs are almost as boring: -0.45, -0.70, -0.44, respectively for groups 1,2,3. So, somewhat you don’t want too much or too little deviation. I guess you can presume that too much deviation might just as much fool the batter when he swings as to make it fall out of the strike zone and not fool the batter at all.
Cutters: -0.67, -0.67, -1.07. So, alot of deviation has some benefit, but generally speaking, not that much of an effect.
Changeups is where the fun happens: -0.47, -0.94, -1.05. So, you do want some amount of deviation, but again, that’s a general rule that applies to someone like deGrom. Kyle Hendricks for example: -2.78, -2.30, +0.58. So for him, he doesn’t want too much deviation. It’s possible that since he lives on the edge of the strike zone, that too much deviation simply means it’ll fall out of the zone. Maybe. At this point, this is just speculation on my part.
Finally, sinkers: +0.09, -0.51, -0.77. Somewhat similar to changeups, but more pronounced: more deviation the better.
Next step
More games needed. At this point, this is just based on a 60-game season. We’d really like to have 50+ starts for each pitcher to come up with any kind of conclusion. When I look at the individual pitcher data, it’s really all over the place. Using 150 pitches as a cutoff is really pretty low. I’d feel much better with pitchers having at least 500 pitches, or even 1000 pitches. In any case, we’ll get there eventually. This is just the first volley.