Pitching research has advanced by leaps and bounds over the last few years — Stuff+ needs to adapt.
A few years ago, I introduced Stuff+ on these digital pages to help power the pitcher analysis in the series previews and elsewhere. It was an attempt to measure the raw stuff of a pitcher’s arsenal using the physical characteristics of their pitches. Back in 2019, this type of pitch evaluation was still a new frontier. Eno Sarris, Jeff Zimmerman, and a handful of others had dipped their toes into creating arsenal scores like this.
Fast forward two years and the amount of research into what makes individual pitches successful has grown exponentially. The biggest leap forward has been a much deeper understanding of how the spin imparted on a given pitch affects its flight path. In 2020, MLB updated it’s tracking system from Trackman to Hawk-Eye, giving us much more detailed information about how a pitch travels from the hand to the plate. The result has been a huge leap forward in pitch research.
It’s almost comical to look back on the components that were included and the way Stuff+ was calculated. I was using the data that was publicly available at the time, but as is the case with many things in baseball, we just didn’t fully understand the minutiae of the game. With more public data in hand, I figured it was time to update the way I calculate my Stuff+ scores to reflect the current research.
Spinning Stuff+ 2.0
The biggest change to Stuff+ is the way I approach spin rate. Back in 2019, we knew that spin rate was highly correlated with velocity and movement and, generally, high-spin pitches resulted in high whiff rates. With Hawk-Eye cameras installed across baseball, we now have the ability to directly measure the spin axis of a given pitch. Before 2020, we had been inferring spin axis based on the movement of the pitch, but it turns out that pitches don’t often move like we expect them to based on the rate and direction of their spin when leaving the pitcher’s hand.
This effect has been dubbed “seam-shifted wake” and the research into it has been gaining plenty of steam over the last year or so. Essentially, the idea is that seam-shifted pitches introduce a different type of spin to a pitch that affects the ball’s flight. A pitch’s spin direction will look one way out of the pitcher’s hand but the actual movement of the pitch when it crosses the plate suggests that the original spin direction was different.
There’s still so much we don’t know about this effect and the type of movement it creates, but one thing is apparent, not all spin is created equally. Some pitches rely solely on Magnus spin — back- or top-spin that pushes pitches up or down vertically, think four-seam fastballs and curveballs — while other pitches utilize a combination of Magnus and seam-shifted spin to generate the movement profiles. Tom Tango has done some initial research into which pitch types benefit from the different types of spin. Sinkers, cutters, and sliders in particular seem to have a positive relationship with seam-shifted wake. When we measure spin direction, we can compare the spin direction out of the hand and the inferred spin direction based on the movement of the pitch when it crosses home plate. The deviation between these two measurements is a good proxy for measuring the effect of seam-shifted wake on a pitch.
Baseball Savant has introduced very handy graphics that show this deviation. Here’s the inferred (left) and observed (right) spin direction of Yusei Kikuchi’s arsenal.
My Stuff+ scores now take this new seam-shifted research into account. They’re also using active spin rates — the amount of spin that’s actually contributing to movement rather than raw total spin.
Commanding Stuff+ 2.0
The other change I’ve made to the way I calculate these scores is how I quantify “command.” This was a major drawback of Stuff+ 1.0. It’s extremely difficult to measure a pitcher’s command using publicly available data. The biggest missing piece is intent. We have location data but we just don’t know if a pitcher meant to locate their pitch in any given area based on where it ended up. The best we can do is understand which areas of the strike zone are beneficial for the pitcher and which areas are best to avoid. That’s essentially what Max Bay has done in this short Twitter thread:
An issue that’s often brought up re: Command+ is how plate coordinate data doesn’t capture intent. This is true in a literal sense, but there are clear pitch-type-specific regional preferences. Those regions generally correspond to regions of lower run value (higher xRV+). 1/2 pic.twitter.com/22sB4snf1R
— Max Bay (@choice_fielder) April 9, 2021
Using Statcast’s attack zones, I queried which regions of the zone possess negative run values per pitch type, and adjusted my command component to take into account pitchers who locate in those regions frequently. Essentially, it’s important to throw strikes with fastballs while avoiding the middle of the plate, breaking balls are surprisingly effective up in the zone and obviously effective down, and offspeed pitches must be located down in the zone to be effective.
Since Hawk-Eye cameras were only first used during the 2020 season, the sample I’m working with is relatively small. Once we reach mid-season, I’ll likely update the data with pitches thrown in 2021. For now, I’ll present the Stuff+ scores for the Mariners starting rotation so that it can be used as a reference in the series previews this year. I’ll update the Mariners data monthly so that it can be referenced throughout the season.
Marco’s changeup grades out really well via this updated methodology but the rest of his arsenal is lacking. It really emphasizes the critical role his command plays in his success. If he can’t locate his pitches — like he was struggling to do early this season — his whole approach falls apart.
Yusei Kikuchi has the best raw stuff on the Mariners pitching staff with a four-seamer and cutter that grade out particularly well. He’s also able to locate both of his secondary pitches extremely well. I didn’t list his changeup that he’s sporadically thrown this year (just 26 times), but it could add an interesting wrinkle to his repertoire sitting between his slow slider and hard cutter.
Flexen’s success in Korea was built upon his fastball and his curveball but those two pitches grade out pretty poorly by this metric. Instead, it’s his cutter and changeup that look like they’re carrying his arsenal. One thing to note are his struggles to locate his any of pitches in beneficial regions.
Despite the low Stuff+ score, Justus Sheffield’s slider is the only pitch in his repertoire he can use to get whiffs, and the pitch’s whiff rate is rather average. His other two pitches are used to generate weak contact, but that’s a fine line to walk for a pitcher with a shallow arsenal.
Justin Dunn is the only starting pitcher on the Mariners with a full arsenal of above average pitches. Looking at the raw characteristics of his two breaking balls shows them to be strong pitches on their own, but when taken together, you can see why he needs to work on differentiating them more.