How does release point affect a pitcher’s success?
Probably the least used of all the great new variables that PITCHf/x has given us is the pitcher’s release point. This isn’t a big surprise; one would assume that things like velocity of the fastball or downward movement of the curve would be more important.
But recently, as I have been profiling pitchers, issues with release point keep coming up. Some pitchers have an extremely repeatable delivery which results in very consistent release points. Others vary theirs. Some throw over the top; others choose a sidearm or even submariner delivery.
How do these things affect the pitcher’s results? I know of no study on release points yet, but we finally have the data to determine what approaches tend to work the best at getting major league hitters out.
First, a word of caution. For this study I am using the PITCHf/x data with the ball moved back to 55 feet from home plate. I will call this the release point even though some pitchers likely release the ball several feet back from this. (No pitcher should be releasing the ball after this). PITCHf/x works only while the ball is in the air, so it can’t tell us exactly at what point the pitcher released it. But it can get close.
Variability in release point
When I profiled Phil Hughes, I noted that he had an extremely consistent release point and wondered if that might be an issue for him. The theory goes that if the batter knows exactly where the ball is coming from, he can key on that location and quickly start identifying the type of pitch and location. On the opposite end of the spectrum is a pitcher like Bronson Arroyo, who will randomly drop his arm slot, disguising where the ball is coming from. Here are some nice pictures showing just that.
Bronson Arroyo’s over the top delivery. Reds vs. Astros, April 23, 2008 (Icon/SMI) | Bronson Arroyo’s sidearm delivery. Reds vs. Astros, April 23, 2008 (Icon/SMI) |
Ignoring any potential injury concerns of a very variable release point, how much does that help or hurt the pitcher? Here is another plot that I initially made for the Hughes article, now updated.
This plot shows every pitcher who has thrown at least 100 pitches that have been tracked by PITCHf/x this year and the standard deviation of their release point. Standard deviation is a mathematical tool for quantifying the spread of the data. If a pitcher throws half his pitches over the top and the other half sidearmed, he is going to have a large standard deviation in his release point. You can see that this plot peaks around one inch with a very long tail to over half a foot.
Now that we know what this distribution looks like, we can plot it against success to identify a pattern. Here, success will be measured in runs100.
Now that is a mess. Runs100 is good for the pitcher the more negative you go, and the league average here is not at zero. This is because runs100 uses linear weights that have been tuned to previous years, and scoring is down a bit this year. That isn’t what we care about here, however. We just want to see what happens to runs100 as a pitcher increases the variability in his release point. As you can see, no pattern is easily discernible and if you run a correlation between these two variables it comes out at -0.06, which is considered to be not correlated.
So this is a somewhat counterintuitive result. I would have expected that the higher variability in release point, the more success a pitcher would have, assuming that would cost the hitter precious time picking up the ball. It seems that major league hitters, at least, are so finely tuned that they don’t care how much a pitcher is changing his release point—they are picking the ball up just fine. It is also possible that pitchers who have variable release points lose some of their control, which is showing up here.
What about the actual location that the ball is being released from?
Horizontal release point
You often hear a commentator talking about how tough a pitcher is because of where he stands on the rubber and where his release point is. This is especially true in a lefty-lefty matchup, where a LOOGY who throws sidearm has been brought in to face a tough lefty. The commentator will say that the ball looks like it is coming from first base and it appears it is going to hit the batter before bending over for a strike. While not a LOOGY, B.J. Ryan is a great example of this. He stands as close to first base as possible and throws nearly sidearm. Here is a great picture that illustrates this.
B.J. Ryan is at the extreme for horizontal release points. (Icon/SMI) |
You can see his back foot at the extreme edge of the mound and his lowered arm slot perfectly here. In fact, PITCHf/x says he is releasing the ball from nearly four feet right of the center of home plate from the catcher’s view. In fact, this is the second most extreme release point for pitchers who have thrown at least 100 tracked pitches. Brewers rookie Mitch Stetter owns that distinction at 4.25 feet! How much does this helps these guys?
To determine that, we will now plot average horizontal release point by runs100 for left- and right-handed batters. Again, my expectation here is that pitchers at the extreme will be excellent against batters swinging from the same side and poor against batters with the platoon advantage. Starting with left-handed batters:
This plot looks like a double-barreled buckshot was unloaded on a piece of paper. Our view is again from the catcher’s perspective, so the cluster on the right is from left-handed pitchers and the cluster on the left is from right-handed pitchers. If a pitcher stands on the opposite part of the mound from his handedness, he will likely show up close to the middle, but if he throws sidearm or stands on the extreme edge, he will likely show up on the edges.
You can see right off that left-handed pitchers tend to do better than right-handed pitchers against left-handed batters. No surprise there. What is surprising is there appears to be no correlation between the horizontal release point and success within each cluster. The lefties who are standing on the extreme first base side aren’t getting any more advantage than the lefties who stand on the third base side. What about right handers?
Same deal here, though there is a hint that extreme lefties are being knocked around by right-handed batters. This correlation is 0.11, which is barely in the small correlation range. All other correlations are below that in the not-correlated range. Again, a somewhat counterintuitive result. I wanted to show next how vertical release point affected things, but ran into some difficultly as, in general, pitchers who are taller throw faster and faster pitches are better pitches. My guess now is that once the speed differential is properly removed, vertical release points wouldn’t matter much to hitters either.
Conclusions
While you will hear from time to time that a pitcher is extra difficult because he varies his release point or because of an extreme release point, it doesn’t appear that either of these things makes a difference at the major league level. While this not the result I expected, it does have some interesting implications. If you are a young pitcher honing his craft with big league aspirations, don’t mess around with dropping down sidearm just to show the hitters something different. It might work at your current level, but it will be wasted if you reach the big leagues. Work on your mechanics and develop a very repeatable delivery even if that means a consistent spot for hitters to locate the ball.
Second, choose an arm angle and a spot on the mound that you are comfortable with and don’t mess with it. If you are a lefty who might profile as a LOOGY, don’t stand on the extreme first base side and throw sidearm in hopes of having better success against lefties. Just keep the style you are comfortable with; it’s what will work the best.
I contribute for Rotographs and i’m working on a z-contact-related post, and I’m looking for updated RMS data – ideally a release point consistency between pitches leaderboard and it seemed like you had a data set for this post…i’d like to learn where i can pull updated data with operational ease of use in mind. Thanks much!