A couple of days ago, I looked at which pitch-tracking measurements most affect the swinging-strike, strikeout, and ground ball rates of curveballs. You can read it here. Why did I choose those variables? Well, they are all factors that pitchers have shown the ability to control. Pitchers who allow more ground balls tend to be more effective. Swinging-strike rates correlate pretty well with strikeouts– which is probably the easiest way for a pitcher to succeed. And strikeouts are strikeouts.
In short, I found that velocity matters. All else equal, a higher velocity on a curveball leads to more swings-and-misses, strikeouts, and ground balls.
Today I will expand my analysis to changeups. The methodology is similar, but not exactly the same. I’m looking at whiff per swing percentage instead of swinging-strike percentage. They are very similar, except the former is on a per swing basis while the latter is on a per pitch basis. I’ll get to the “why” in a second. Again, I’m looking at ground ball rates too (I’m not looking at K%. Swinging-strike% correlates well with K% so doing both is not entirely necessary).
As for the “why”: I realized in my analysis of curveballs that I was including pitchers’ swinging-strike%, K%, and GB% as a whole as opposed to looking at those rates solely from curveballs– which is very different and may have led to some inaccurate results. So, today, I’m including the two rate stats calculated only from when the pitcher threw a changeup. It was much easier to find whiff% grouped by pitch type than it was to find swinging-strike%, so that is why I made the change.
The pitch-tracking measurements included in the curveball analysis remain here– velocity (aka start_speed), spin rate, and horizontal and vertical movement. I’ve also included a variable that equals the velocity difference between the pitchers’ changeup and fastball. Out of the pitchers hand, a fastball and change should look identical. Part of a change’s effectiveness (I think) lies in it’s deceptively slower speed. The greater the velocity difference the better (again, I think).
First up is whiff percentage. Again, this 258-pitcher sample includes pitchers, since 2014, that have faced at least 150 batters and thrown at least 200 changeups and 5 fastballs (fastball velocity stabilizes almost immediately). Pitchers’ whiff percentage on a changeup is described as a function of velocity, spin, horizontal and vertical movement, and fb/ch velocity difference. Results are shown here:
You’ll see that velocity, vertical movement, and velocity difference are significant. Meaning, they all contribute to a swing-and-miss. Velocity difference matters the most, then velocity and vertical movement matter to a lesser extent. Some interpretation:
- All else equal, a pitcher with a greater fb/ch difference will generate more swings and misses
- For pitchers with an identical fb/ch velocity difference, a faster changeup will result in more swings and misses
- For pitchers with an identical velocity difference and changeup velocity, more vertical drop will result in more swings and misses
If visuals are more your thing, here is the relationship between whiff percentage and fb/ch velocity difference:
You can clearly see the positive relationship up until around a 11 mph difference. After that, there aren’t really enough data points to make any conclusions. Interestingly, Felix Hernandez is able to generate an above-average whiff percentage (36%, mean= 30%) with almost no difference between his fastball and changeup velocity (3.5 mph). How? He gets the third-most drop on his changeup and has the seventh-highest changeup velocity in the sample.
We find a similar story when looking at ground ball rate. Velocity, velocity difference, and vertical drop all matter. Though, as you can see by the negative coefficient, less velocity difference actually leads to more ground balls. This is interesting. It makes me pause and think about “why” for a second. A greater velocity difference leads to the batter’s bat head crossing through the hitting zone before the ball gets there. Thus, a swing and miss. A smaller velocity difference also results in the bat head crossing through the hitting zone prior to the ideal time, but this time the batter is able to make contact because the difference is less. However, the contact is usually weak (because the ball is hitting the end of the bat) and often times, a ground ball is the result. Two different, yet effective, techniques.
For a visual piece, I will focus on what has the greatest effect on causing ground balls– vertical movement. An additional inch of vertical drop increases ground ball rate by 2 percentage points. The relationship is shown here (a smaller vert_mov value means more vertical drop).
Compared to previous plots, this relationship is rather clean and obvious. One outlier though, is Rafael Montero, a pitcher in the Mets’ organization who has bounced around between AAA and the majors over the last three seasons. His changeup is hit in the air 75% of the time, the highest in the sample. When you look at his peripherals though, this doesn’t really make sense. He’s right around average in velocity, velocity difference, and vertical movement. The model actually predicts his ground ball rate at 51%, which is above-average (mean= 49%). And why is Montero’s changeup hit in the air more than it theoretically should? My thought is that location could be a factor. But that’s a question for another time.
To recap:
