A haphazard dive into Baseball Savant data.
Since Statcast started tracking spin rates of pitches in 2015, discussions about spin rates have become mainstream, with even regional and national broadcasts making reference to them with some regularity. The stat itself came into the spotlight in 2021 in the aftermath of MLB’s crackdown on pitchers’ use of sticky substances. A good portion of fans looked for which pitchers lost the most spin on their pitches, and even the New York Times and Washington Post got in on the discussion.
The effects of high spin rates are pretty easy to explain and I’m sure most VEB readers are familiar with them already. Four-seam fastballs (especially ones thrown with a high arm slot) carry higher through the zone as the increased Magnus effects offer extra resistance to gravity. Curveballs will drop more severely, sliders will break to the pitcher’s glove side more fiercely, and sinkers (depending on the grip and arm slot) will see more arm-side action. With all the extra movement from higher spin rates, it would be logical to assume that hitters will have a harder time squaring up pitches as spin rates increase. At least, that was my impression of spin rates as they became more mainstream. Curious about if the trends show up in the Statcast data, I decided to plot the data for common pitches from the 2021 season, comparing pitchers’ average spin rates in RPM on each pitch to their opponents’ wOBA and average exit velocities in MPH. Each data point represents a single player who threw that type of pitch to at least fifty batters in 2021.
(A quick note on the plots below: Baseball Savant was giving me difficulty with the graphs they provided in the searches of their database, so I decided to export the data and plot it using R. My skills with R are still pretty limited, so the graphs are rudimentary but hopefully they get the point across.)
Looking at these plots, there doesn’t seem to be any linear correlation between these factors. When applying linear regression, the highest correlations kick back values of 0.02 to 0.03. Exit velocities also seem to be distributed randomly throughout each plot. So are the effects of spin rates overstated? Are teams wasting their time and money researching this topic? I think, despite the plots above, that the obvious answer is no. Between the cost of the technology needed to analyze pitches and the time it takes to optimize data collection, run analysis, and provide feedback on this topic, I find it hard to believe that MLB teams would be wasting resources on emphasizing it.
One obvious conclusion is that spin rate isn’t the end-all-be-all of pitch quality. The qualifiers mentioned above, such as arm slot and pitch grip, affect how that spin is applied, which is another essential aspect in analyzing pitch movement. A good example of this is Giovanny Gallegos’ slider. Thrown from a three-quarters arm slot, Gallegos’ slider has a considerable amount of vertical drop to it, while sliders thrown from a sidearm motion, such as Andrew Miller’s, have much more of a flat, sweeping action across the zone. The difference in the vertical drop is also part of the reason that Andrew Miller has been so susceptible to right-handed hitters since pitches with vertical drop are tougher to hit for hitters on both sides of the plate compared to pitches that stay closer to the same plane.
The other factor that may be scrambling the data is that I may be looking at the wrong variables. Though spin rate doesn’t correlate with wOBA or average exit velocity, it may have a correlation with other events. For example, according to the Statcast data, four-seam fastballs with spin rates that fall at the low and high ends of the distribution seem to correlate with fewer home runs given up. This helps explain why Cubs’ starter Kyle Hendricks, despite being in the bottom one percent in fastball spin, has carved out a nice career for himself. Additionally, it’s now commonly said among the sabermetric community that higher deviation from the most common spin rates (at least, for fastballs) is more of a factor than how high the spin rate is itself, which is borne out by that data.
Spin rate is a tough topic to analyze given the number of variables that need to be taken into account. Everything from the release point of the pitch, to the grip, to pitch velocity is going to affect the results seen above and the analysis will probably need to control for more of these factors before any meaningful results can be found. Additionally, wOBA is probably a bad stat to look for correlations in given its high variance in small sample sizes (which is why exit velocity was also included above). This is something I’m more interested in exploring and writing about if there’s a demand for it on VEB, and any feedback in refining the process is more than welcome. Once the analysis (and my skills in R) are more refined, I’d also like to revisit this data and see where Cardinals’ pitchers fall in these datasets. For those interested in more informed reading on the topic in the meantime, I’ll once again refer to Driveline and their studies on the topic, since that will be the guide for refining this analysis in the future.