Michael Fulmer's unique slider could be late-inning feature in Cubs bullpen

Remember a few years ago when trade rumors between the Cubs and the Tigers circulated around Michael Fulmer? Fulmer, who had just been named the American League Rookie of the Year in 2016, was viewed as a cost-controlled workhorse who could yield a return of Kyle Schwarber. However, just three years later, Tommy John surgery kept him out for the entire 2019 season and effectively ended his career as a starter.

Fulmer, now a reliever who reportedly agreed to a deal with the Cubs on Friday, has had a career renaissance over the last two years. In 133 innings between the 2021 and 2022 seasons, he impressed with roughly a 3.5 FIP and averaged nearly one strikeout per inning, thanks to a gyro-esque slider he throws more than any other pitch.

When Fulmer won his 2016 Rookie of the Year award, he utilized a slider a little more than once every four pitches. But as a reliever last season, he threw a slider more than three times for every five pitches, which was among the highest frequencies in Major League Baseball.

You can understand why Fulmer threw his slider so often. Of 520 qualified pitchers last season, Fulmer ranked 20th in slider run value — approximately plus-two standard deviations. So, what makes his gyro slider so good?

A gyro slider is thrown with a similar grip to a slider or cutter, but the pitcher applies extra pressure with the middle finger to create a greater amount of spin on the ball. The increased spin creates a more pronounced vertical break and a tighter, more vertical spin axis. When thrown over 90 mph, like Fulmer’s, the pitch has extremely late and sharp break.

Michael Fulmer is an interesting signing.

He was successful because he threw >60% sliders, which has gyro action with fast velo. pic.twitter.com/x1IgELacrl

— Brendan Miller (@brendan_cubs) February 10, 2023

Few pitchers are even capable of throwing Fulmer’s gyro-bullet slider. You can visualize his slider in data format below. The x-axis (horizontal axis) shows Fulmer’s spin efficiency (roughly 30 percent), and the y-axis (vertical axis) highlights Fulmer’s 90 mph velocity. Every hexagon is just a bucket of pitches from MLB. The cross-hairs are Fulmer.

You can see that few pitchers throw sliders over 90 mph, and even fewer throw it with a spin efficiency under 30 percent (note: you want lower spin efficiency for gyro sliders). However, the pitchers that do throw these sliders typically have “Stuff” grades near 70 on the 80 scale, or about two standard deviations better than league average.

Fulmer’s slider, though, rated a 50 on the 80 scale, or league average. So, you might ask the question: Why does his stuff rate league average if its value was greater than 95 percent of MLB?

Now, it’s important to understand how these stuff graders work. They are extremely powerful and useful tools that substantially aid front offices, coaches, players and scouts.

The graders are created by machine learning algorithms that take dozens of features (horizontal break, spin rate, etc.) and attempt to predict run probability on a pitch-by-pitch basis. Cameron Grove’s model is trained on 80 percent of Statcast data (2015-2020 seasons) and tested on the remaining 20 percent. The model is supervised by run probability, which means it will try to find the thresholds that best align with run probability.

The computer will create dozens to hundreds of models and then find the one with the best accuracy, and then it will try to break the model by removing features one-by-one to see which feature, when removed, reduces accuracy the most — these are what are considered the most important features. Across multiple model types, often the important features are velocity and break difference from the fastball. Here’s an in-depth breakdown of Grove’s model.

Graders are extremely accurate. For example, Grove’s model’s predicted run value vs. actual run value were almost perfectly aligned. But despite its near perfection, there is some variance at the extreme ends of the line. This means that despite its impressive accuracy, other variables are at play. In Fulmer’s case, it’s within the realm of possibility that his 50 (out of 80) slider grade is not truly that, because its defining features are rare (i.e., less data), which won’t massively influence the algorithm’s decision making. This is all conjecture, though.

Regardless, the Fulmer signing is interesting because he has a unicorn slider. It’s by no means one of the nastiest pitches in the league, but it’s among the most effective sliders. The question is whether he can command his package of pitches — something he has struggled with at times (3.96 BB/9 last year) — and induce enough strikeouts without a put-away pitch.

In a dream scenario, if the Cubs were to add another breaker to his repertoire (e.g., a slider with more sweep), then he could absolutely be a closer, because he would then have that put-away pitch. That’s easier said than done, of course, and it’s likely he remains in his expected role, but that’s why it’s a “dream scenario.”