There’s no question that we currently live in an age of big data, with a surplus of information and analytical thinking stretching from the political world to the bar. Sports are no exception, and baseball, given its individualistic nature and easily tracked moments in a controlled environment, has long been seen as ground zero for data analysis in sports. But as earned run average (ERA) has given way to fielding independent pitching (FIP), and value over replacement player (VORP) has ceded to wins above replacement (WAR), baseball data and its increasingly available nature consistently reminds analysts that baseball is like an ever expanding onion, where it’s impossible to comb through all of the layers without new ones intruding. In this vein, Statcast data (which became publicly available at the beginning of the 2015 season) simultaneously combs through the onion with measures like exit velocity and launch angle, while also adding additional nuanced layers in the form of measurements not visible to the naked eye (i.e. effective velocity, release point, etc.). Spin rate is one of the latter forms of Statcast measurements and early indications show that spin rate may hold the key to quantifying pitch movement, which if tracked and interpreted correctly could join velocity as the two primary attributes teams look for when evaluating pitchers.
Generally speaking there are three overarching skills that a pitcher can employ to be effective:
History has shown us that pitchers do not need to have all of these skills to excel (very few can claim to have all), and time and time again baseball has shown that different pitchers can have sustained success against hitters despite possessing distinctly different styles. After all, Greg Maddux and Roger Clemens were the two greatest pitchers of their era and the only thing either of them had in common was that they were both right handed. Maddux was such a legend in the art of pitch control and pitch movement, that his lack of velocity was an afterthought. But during the prime of Maddux’s career the only one of the aforementioned skills that the public could measure/analyze was his below average velocity; fans could see Maddux’s greatness, but it wasn’t quantifiable (at least on a pitch by pitch basis, Maddux still won four Cy Youngs so don’t feel too sorry for him). With the release of PitchFx data in 2008, amateur analysts now had the ability to plot the location of every pitch thrown, and now with Statcast’s spin rate pitch movement may be next. The goal of this analysis is to build off of Mike Petriello’s (of mlb.com) previous analysis regarding the relationship between spin rate vs. velocity. In part one herein I will examine the relationship between spin rate vs. pitch velocity for individual pitch types (i.e. four seam fastball, sinker, curveball, etc.), specifically what combinations of these two variables lead to higher swinging strike probabilities. In part two to be released later next week I will conduct the same analysis, only this time I will examine which combinations of spin rate vs. pitch velocity lead to higher soft contact (i.e groundball/pop up) probabilities.
In the plot above I’ve utilized Statcast data from 2015-2016 (regular season only) and plotted the average spin rate and average velocity (minimum one hundred pitches thrown) for each qualifying offering by pitcher. As you can see it’s possible to make broad inferences from the plot above (i.e. four seam fastballs on average posses the greatest average velocity, curveballs possess on average the greatest average spin rate, etc.), but there are too many data points on this plot to make specific conclusions. As a result I’ve broken out below each data point into one of three categories: Fastballs, Breaking Balls, and Offspeed Pitches.
Even when broken out it’s hard to see any type of linear trend between velocity vs. spin rate for any of the above pitches. The regression analysis (which also includes n = sample sizes) below confirms this with the highest r2 belonging to the cutter at 0.145.
|Pitch Type||r 2|
When you think about this it shouldn’t be a total surprise, after all pitch types are a somewhat arbitrary designation as different pitchers possess repertoires and varying delivery angles that don’t fit into an easily categorized boxes. In other words a four seam fastball looks very different when it’s delivered by Koji Uehara vs. Max Scherzer even though it’s classified as the same pitch type, the same could be applied sinkers delivered by Noah Syndegaard vs. Kyle Hendricks, curveballs delivered by Rich Hill vs. Michael Wacha, etc.
Keeping these differences in mind I set out to find the relationship between spin rate (rpm) and velocity (mph) for swinging strikes for 2015-2016. To do so I compiled a dataset of all Statcast tracked pitches thrown (taken from baseballsavant.com) during the 2015 and 2016 regular seasons (n=1,286,980). Then I filtered each pitch type I wanted to examine (i.e. Four Seam Fastball, Curveball, etc.) into separate datasets taken from the master dataset. After that, as I’ve done before on the Exploring Baseball Data with R blog, I utilized a generalized additive model (GAM) to calculate the probability of a swinging strike based on spin rate and velocity for each pitch type. In each of the swinging strike probabilities below I’ve plotted the mean spin rate, and one standard deviation above and below the mean spin rate. The lightest blue lines represent one standard deviation above the mean for that pitch type, while the darkest blue lines represent one standard deviation below the mean for that pitch type. In the table below I’ve listed out each pitch type’s mean spin rate, standard deviation of spin rate, and sample size.
|Pitch Type||Mean Spin Rate||Standard Deviation Spin Rate||Sample Size|
|Four Seam Fastball||2249.4||193.2||494,233|
Fastball Swinging Strike Probabilities: For each of the three different types of fastballs the greater the spin rate, the higher the probability of a swinging strike. This isn’t surprising for four seam fastballs, as a greater spin rate leads to a rising effect in which the ball crosses the plate at a higher point than where the batter expects. The same can be said about cutters (aka cut-fastballs), as Kenley Jansen (the heir apparent to Mariano Rivera) largely rode his cutter (which possessed an average spin rate of 2,562 rpm, nearly an entire standard deviation above the mean cutter spin rate) to an approximately forty percent strikeout rate during the 2015-2016 seasons. Of the pitches I ran the swinging strike probability GAM for, cutters had the greatest disparity between high spin rate and low spin rate. Simply put it doesn’t matter how hard you throw your cutter, if it doesn’t have spin it faces an uphill battle in order to be an effective swing and miss pitch. The sinker (aka two seam fastballs) results surprised me a little as it’s a pitch that designed to drop (or sink) in the strike zone, and one would think that a higher spin rate (with a similar rise effect that a four seam fastball has) would prevent this from happening. However, the probability of a swinging strike for sinkers does not reach the probabilities for four seam fastballs or cutters. This makes sense as pitchers who utilize sinkers often throw them to engender (what they hope to be) weak contact, and it has historically not been seen as a “put away” pitch for the majority of pitchers.
Breaking Ball Swinging Strike Probabilities: Breaking balls had some of the lowest r2 values, but for each of the three pitches plotted above (Slider, Curveball, Knuckle Curve) there generally seems to be more swings and misses if you’re able to get more spin on the ball. This is most apparent when looking at the slider swinging strike probabilities above; pitchers who can combine high spin rates with above average velocity can possess potentially devastating sliders. This was most apparent during the 2016 playoffs as the Indians nearly rode the sliders of Corey Kluber, Andrew Miller, and (to a lesser extent) Bryan Shaw to a World Series victory. While not as drastic as sliders with high spin rates, curveballs with higher spin rates also have a higher probability of getting swinging strikes. In 2016 Rich Hill and Seth Lugo in particular were able to harness their high spin rate curveballs to become valuable contributors to NL playoff teams. It will be interesting to see if these curveball spin rate vs. swinging strike trends continue as the total number of curveballs thrown continues to increase in 2017 and beyond.Spin rate does not appear to play a large role in determining the swinging strike probabilities for knuckle curves; in fact the average spin rate for knuckle curves had the lowest swinging strike probability. Of the six breaking/fastball pitches that I ran this GAM on, knuckle curves were the only one that had both the lowest spin rate (one std. deviation below), and the highest spin rate (one std. deviation above) clearly above the mean spin rate in terms of swinging strike probability. Since knuckle curves had the lowest sample size in the swinging strike probability analysis, it is possible that with more knuckle curves to analyze these swinging strike probability results may change.
Offspeed Pitches Swinging Strike Probabilities: Since changeups are highly dependent on a pitcher’s fastball, it makes sense that velocity and arm angle have more of an impact on swinging strike probabilities than spin rate. Case in point, Danny Duffy and Kyle Hendricks both had breakout seasons in 2016 thanks in no small part to getting more swinging strikes via their changeups, Cole Hamels maintained his excellent performance in 2016 even though he continued to depart from his physical prime in large part because of his changeup’s effectiveness, while Noah Syndegaard ascended arguably to best pitcher alive (not named Clayton Kershaw) status in part thanks to his improved changeup. The average changeup spin rates for these four pitchers in 2016 were 2512 rpm, 2118 rpm, 1626 rpm, and 1765 rpm respectively. Similar to sinkers, splitters are most effective at getting swing and misses when there is a lower spin rate. This makes sense since splitters that have too much spin tend to stay up in the strike zone and become very appetizing to big league hitters. This was the case early in the 2016 season when Mashahiro Tanaka’s splitter with a spin rate of 1547 stayed up in the strike zone, and Adam Eaton gladly deposited it into the Yankee Stadium right field bullpen.
This concludes the spin rate and swinging strike probabilities analysis. As you can see, depending on the pitch type, spin rate can either be very important at inducing swinging strikes, or not play a central role at all. It’s important to remember that I’ve only analyzed two seasons of Statcast data since that is all that is currently publicly available. With more data these results could be clarified and expressed with greater confidence, but even if more data was available, new layers of the baseball onion would inevitably appear. And that’s ok, the fun was “always in the chase” so to speak in baseball analysis, unequivocally figuring out each facet of the game would take the joy out of watching it. On that note if you enjoyed this article be sure to check back next week when I tackle the relationship between spin rate and soft contact.