The St. Louis Cardinals made some noise over the weekend when they signed their former-prospect-turned-dominant-reliever-for-other-teams, Luke Gregerson. Cardinals Twitter seemed to mostly like the acquisition, which makes sense given that many of us have been "screaming for years that the missing piece goes by the moniker Luke Gregerson." Well, now the Cardinals have Luke Gregerson.
That being said, we are left to wonder which Luke Gregerson -- who turns 34 next May -- the Cardinals are actually getting. In eight seasons through 2016, Gregerson posted a very good 2.84 ERA supported by a 2.98 FIP and 3.15 xFIP. In 2017, he posted a 4.57 ERA, with a 4.62 FIP and 3.45 xFIP. Craig Edwards of Viva El Birdos expressed some reason for concern about his outlook, but concludes by saying that there "are good reasons to expect a bounce-back performance" in 2018.
Before I delve into my analysis and concerns regarding the signing, I first want to take a step back. Luke Gregerson, over the course of his nine-year MLB career, has made a living on his fantastic slider, or, perhaps more appropriately, his fantastic sliders. Back in 2013, Eno Sarris detailed evidence that Gregerson actually throws three separate, distinct sliders. According to Gregerson himself, he throws one that has a flatter break, one that breaks straight down, and one that breaks somewhere in between. Well, at least he used to, that is.
You can see the visuals in Eno's article, but I've recreated some here using data available at BaseballSavant.com. This way, I can control the scaling and make easier comparisons to what's happened in the years since 2013.
Below are nearly all of Gregerson's pitches classified as sliders in the 2012 and 2013 seasons. Using Tableau, I performed a cluster analysis to create different categories based upon the pitch's release speed, vertical movement, and horizontal movement.
Here, we clearly see three separate sliders. You might even call them three separate pitches altogether. There's obviously a little overlap, but generally speaking, these are clearly distinct from each other. Looking at the summary data for each type of slider adds some clarity, at least in my mind, so I've compiled that below.
Note: A positive vertical break indicates the pitch has some rise, meaning the break looks more flat. A positive horizontal break indicates the pitch is breaking from right to left, from the pitcher's perspective. The break data is measured in feet (I think, as I've never seen it confirmed).
Regardless of the specifics, what we see here is three very different pitches. Gregerson threw a hard slider (orange) that had the least break of the three, but also was a tick-and-a-half faster than the other two. He threw a slider that dove (green), and one that may have looked more like a slurve-y slider (blue). Eno has a few great shots of each type in his old article if visuals are your thing.
In terms of overall value, Gregerson's slider(s) was/were the fifth most valuable among qualified relievers in 2012 and 2013. His slider was great and, subsequently, he was also great.
Then, from 2014 to 2016, Gregerson's slider lost some value. It was still a very good slider, ranking as the 12th most valuable, but it wasn't quite as good as it once was. Looking at the graph below, it's easy to see one major difference - the green cluster no longer looks like it is definitely different. It moved closer to both the orange and blue clusters and, frankly, is less defined overall.
Add in the pitch data:
The blue sliders got a little flatter with more horizontal break. The orange sliders got a little flatter with more horizontal break. The green sliders got a lot flatter with more horizontal break. You didn't need the data to see that the three clusters are converging. You could probably argue that there are really only two clusters at this point with occasional outliers.
Then, in 2017, Gregerson's slider was worth a career worst 0.5 runs above average. It was the first time his slider's value fell below 2.7 runs, and it was nearly a six and a half run drop from his 6.9 run mark in 2016. Why?
Once again, the clusters continued t