Photo Courtesy of Pinstripe Alley
by: Mathew Haines
Dellin Betances is one of the game’s most electric relievers. After maintaining a 15.28 K/9 and 11.6 WAR through 6 seasons, it’s hard to deny the dominate prowess that Betances brings to the mound. Yet, despite the authority he displays on the mound, Betances is often criticized for his lack of control. If you are unaware of these criticisms, take a look at one of the more “outspoken” displays of frustration over Betances’ control here.
Currently, Betances holds a 3.98 BB/9 throughout his entire professional career, nearly a full walk per nine innings more than other top active relievers. However, BB/9 is one of the lowest correlating statistics when looking at an MLB reliever’s wRAA. So the question becomes, what makes Betances such a dominant reliever in the game? Throughout this series of articles, I will be analyzing a variety of pitchers with this question in mind. All aiding into the grand question, who is the most “effectively wild” pitcher in the MLB today? Now, I welcome you to my first installment of “The Effectively Wild Hypothesis.”
An important side note to this article. You will see me referring to pitchers as, “wild,” which would normally be assumed to be a lack of control in a pitcher. Unfortunately, there are not many ways to track how much a pitcher misses his spot. In this instance, I use the word “wild” to refer to the randomness and/or unpredictability of a pitcher.
In order to accurately analyze the anomaly that is Dellin Betances, let’s look at a general overview of his PITCHf/x data acquired by Brooks Baseball. PITCHf/x is a tool created to track the trajectory and movement of every pitch thrown in the MLB. We will look at a variety of data collected, but first, we’ll keep it simple by looking at some tabular data which will break his arsenal down from a fundamental standpoint. This data tracks all of Betances’ pitches thrown throughout his career.
The following table is a breakdown of the outcomes induced on each pitch:
Betances primarily relies on his fastball and curveball, which make up 99.75% of pitches he has thrown in his career. It is important to note that PitchF/X likes to combine curveballs and sliders into one category. This is important when transitioning to Baseball Savant data which classifies the two pitches separately, as will been seen later in the article.
The first thing I notice is his impressive whiff% on both his breaking balls and fastballs. These are Dellin’s primary pitches and have resulted in an astounding 1,007 whiffs throughout his MLB career. While this data is nice to look at and gives a general idea of how Betances attacks his hitters, it fails to give us a good indicator of how “wild” Betances’ pitches tend to be especially on a pitch-by-pitch basis.
First, let’s compare some of these stats to some league wide data.
Unlike most top relievers in the game, Betances favors his breaking pitches more than his fastball. This in turn will suit to raise his variability, for his continuous use of breaking pitches will alter his release point, pitch movement, and velo. Another notable fact about his Betances’ fastball: it’s fast, like, really fast. He currently sits with an average fastball velocity of 94.2, among the 99th percentile among MLB pitchers.
When analyzing the movement on his pitches, we can add a few more metrics to this data. In the table below, I have provided some easy-to-read data that average his pitch break as well as the rate of change (ROC) of the pitches on an X-Y plane. What to look for here is not only how much a pitch breaks or how quickly the pitch gets to the plate, but rather how quickly a batter must adjust to each pitch in order to track it properly.
In the data we see a great deal of horizontal movement, but more importantly, see a large amount of vertical break in every one of Betances’ pitches. This is often overlooked by the casual viewer, for we often digest horizontal break easiest and do not digest the concept of vertical movement well. The vertical movement of a baseball has little to do with tangible movement, but rather the effect of the magnus force has on the ball which changes its trajectory from its hypothetical, dragless path.
Introduction to the Model
For this project, I will be looking at all 1,134 pitches thrown by Dellin Betances in 2018. The sample data can be accessed here via Baseball Savant. To determine how “random” Betances’ pitches are, I will be using a euclidean distance formula to determine how “different” each pitch is from the one before. For those unfamiliar, euclidean distance formulas are essentially ways to find out how “different” or “alike” two things are through using a multi-dimensional mathematical formula. At the most basic level, this is done through “graphing” two items, resembled by a point, each on different planes, then finding the distance between these two items.
First, it can be beneficial to visualize the sample data, so we know just what we’re working with. In this first plot, we will be looking solely at location of the pitches. The following plot lays out the pitch locations of each pitch and color-codes them by pitch type.
Here we see that Betances does not seem to have much of a difference in how he decides to locate his pitches. While this may not seem like a good thing from a pitching coach’s perspective, this is precisely what we are here to analyze. With pitching irregularity like this, is a batter really able to predict where one of Dellin’s pitches are going to end up?
In order to break down Dellin’s pitch location’s a bit more, I have created a density graph of each pitch type in order to see if there is some change in location (for these scatter plots are often hard to digest due to the shear amount of information on them).
As seen, there is a little bit of variation in these graphs. At first glance, it appears that Dellin is most comfortable with his fastball and knuckle-curve, as these pitch types have the lowest variances among the three. Likewise, Dellin is not afraid to pound the fastball down the middle, as shown by this area being the most dense. However, both of his primary breaking balls tend to locate more low and away for a right-handed hitter. This could be evidence of Betances utilizing pitch tunneling, doing what is natural for a RHP when throwing breaking balls, or executing his game-plan, but that’s a subject for a different article. Just for comparison reason, I have included a density graph for a reliever known for better command as well as superb strikeout power, Chad Green.
As shown, there is much more evidence of a specific location for each pitch, fastball being down the middle, splitter down and in to a righty, and slider down and away for a righty. It is important to note the difference in these two sets density graphs. While both pitchers are effective, they both use use their pitches differently to get batters out.
Now that the data has been laid out, it’s time to ask, how do we use it to analyze, “The Effectively Wild Hypothesis?” For this model, I ran a euclidean distance formula that measures how “different” one pitch is from the pitch that came before. To do this, I took 6 primary variables, and then I broke it down into a variety of categories to further analyze Betances as a pitcher. This formula takes the break of each ball from its release to when it cross the plate and compares it to the break and location of the previous pitch . To adjust for the varying speeds of the baseball, one of the most important variables in the variation of pitches, I structured the planes of the euclidean model to coincide with the time the ball takes to reach the plate from the time it was released. I call this stat, Pitcher Variability Rating (PVR). On average, Betances has a 17.26 PVR, meaning that when accounting for the average pitch from Dellin Betances, a batter must account for approximately 17 inches in difference between pitch location, arm slot, and timing. Yet, this is not the whole story. In this next table, you will see the breakdown of Betances’ PVR in a variety of situations.
So what can be concluded from this data? Most notably, Betances’ first pitch of the at bat varies one of the highest amounts from the previous pitch. This is likely due to his “out pitch” being much different than his “opening pitch”. After that, Betance’s 1-0 pitch tends to vary the most from his previous pitch. The inner psychology of Betances’ thoughts are not something I can speculate on, but it appears that he gives batter a varying look early in the count, quickly throwing their perception off. As we continue from count to count, we see that Betances remains fairly consistent with his variation of pitches. While some may call this being “wild,” for there is little trend in location, movement, and location of each pitch. In fact, when looking at these splits of the data, it becomes harder and harder to predict how much (or little) variation in pitches Betances will throw at his hitters based solely on situation. This constant variation in turn forces hitters to constantly make adjustments from pitch to pitch by altering his arm slot, pitch type, and velocity in an unpredictable way. Combine this with a fastball velocity in the 99th percentile of all MLB pitchers, and hitters have quite the task ahead of them if they are to be successful against Betances.
In thinking about the data presented in this article, you might be wondering, how can we use PVR in the future? While one case study does not prove/disprove the “Effectively Wild Hypothesis,” it gives us a measure of how to track a pitchers’ variability on a pitch by pitch basis. Using these patterns, baseball organizations will not only find out in which situations pitchers are lacking, but they offer a new level to scouting reports for hitters when preparing for opposing pitchers.