New Pitch Uniqueness, Pt. 2: The Slambio (and a Ghost Fork update)

This young season has already introduced us to a few unique offerings. Brent Honeywell Jr. throws a true screwball. Kodai Senga throws a ghost fork. But one fascinating pitch has flown somewhat under the radar: Ian Hamilton’s slambio.

Maybe it’s because of the right-hander’s lack of a pedigree, or his status as a non-roster invitee during spring training. After all, Hamilton is a 27-year-old reliever who struggled through injuries and ineffectiveness over the past four years. At the same time, he looked like a find as recently as 2018, when he pitched to a 1.74 ERA and 2.44 FIP across 51.2 innings between Double- and Triple-A; he even averaged 96.7 mph on his heater in a brief eight-inning callup. The next season, he seemed poised to play an integral part in the White Sox bullpen, but he was struck by a foul ball while rehabbing separate shoulder and head issues stemming from a car accident. His poor luck nearly brought his career to an end, but he finally began to feel back to normal this offseason.

If you didn’t know about Hamilton before, I’ll be the first to tell you that he has been a joy to watch this season, not only because of his unique offering but also his comeback story, parlaying his rediscovered health into a spot in the Yankees’ pen, where he’s found early-season success with a resurgent fastball (averaging 95.4) and the slambio. The latter pitch has been nothing short of excellent thus far: a ludicrous 29.9% swinging-strike rate and worth 3.8 runs, which rank third and tied for fourth, respectively, among the 85 sliders thrown at least 50 times this year (as of Saturday night). The pitch’s unusually high rate of called strikes, 15.6%, given its whiffiness, also places it second among the 85 sliders in CSW%.

But is it really a slider? The name “slambio” comes from a combination of “slider” and the Spanish word for changeup or change, “cambio.” Yet in truth, the pitch defies classification. There were four different pitch types among its top six comps from last year’s set of 1,323 pitches that were thrown at least 100 times:

Zoom out a little further, and you’d find that a cutter — Jackson Tetreault’s — ranks as the 19th-closest comp to the slambio. But closest on what terms?

Last week, to assess the ghost fork’s distinctiveness, I used Euclidean distance. It’s the technique I used for the slambio as well, and the one relied upon in algebra to find the distance between two coordinate pairs — it depends on the Pythagorean Theorem, envisioning the line segment connecting the pairs as the hypotenuse of a triangle. This strategy can be used for vectors with more than two variables as well; in the same manner, you just need to take the sum of the squared differences of each pair of values after normalizing them to be on the same scale.

The variables I chose were horizontal and vertical movement (inches), velocity (mph), spin rate (rpm), and spin axis (degrees). I’ve been refining my model since the Senga piece, but in doing so, I discovered an error in my normalization process. Specifically, horizontal movement was weighted improperly; the mean for righties is negative, and I had it in as positive. This didn’t change my bottom line; I still found that pitches with larger mean-of-closest-five Euclidean distances (i.e., pitches that were more distinct) tended to have higher swinging-strike and called-strike rates, results that withstood a control for zone rate. Further, I still found that splitters and changeups with larger mean-of-closest-five Euclidean distances tended to have higher swinging-strike rates, as did those that were more similar to the ghost fork. Similarity to the ghost fork meant fewer called strikes, but the offering is meant as more of a chase pitch anyways, so this might mean the pitch is just doing its job.

The one significant change wrought by the new model, also due in part to having more ghost forks in the dataset, was that the closest comps for the pitch were different. This time, they were all splitters (Mark Leiter Jr. throws a split-change), making the ghost fork’s performance relative to that category of pitches all the more important:

Two of these, García’s and Leiter’s offerings, were previous top comps. The rest are new, and they tell an even whiffier tale of what the ghost fork’s future holds. But despite its more clear fit in the splitter category, the refined model actually paints the ghost fork as even more distinct. Previously, it was 92nd in the dataset of 1,323; now, it places 53rd.

When it came to the slambio, it was harder to compare it to just one pitch type. It was most similar to the average cutter on the horizontal plane and in terms of spin axis, but it lined up more closely with the average splitter in terms of drop, velocity, and spin rate:

Essentially, this is because Hamilton grips the baseball about the seams in the same way that pitchers would grip a typical cutter. Further, he often suppinates his wrist like pitchers usually do to get cutting action. Here’s a good look at it when he suppinates to the max:

Ian Hamilton's "slambio." Changeup grip and spin, slider intent and movement. pic.twitter.com/NMyeUpkMkY

— Lucas (@DBITLefty) April 7, 2023

Yet as you could tell from the clip, Hamilton holds the ball with all of his fingers on it like a changeup, and sometimes he even pronates his wrist to get more changeup- or splitter-like action:

Ian Hamilton's "slambio" is FILTHY ?#YANKSonYES pic.twitter.com/BENhi2nBaP

— YES Network (@YESNetwork) April 4, 2023

On the whole, changeup-like and slider-like slambios put together, the refined model saw the pitch as the eighth-most distinct. Tyler Rogers still owned two of the top five spots with his four-seamer and slider due to his unique release point; Devin Williams‘ airbender, Eli Morgan‘s slow change, and Dustin May‘s nasty curve marked new additions. Logan Webb’s diving changeup and Leiter’s split-change also squeaked in ahead of the slambio, but Hamilton’s offering joins a small number of pitches that truly defy categorization, providing further evidence of its uniqueness beyond just its closest comps.

Is it really fair, though, to treat the two flavors of slambio as one pitch? Consider the following:

Aside from the massive amount of seam-shifted wake that the slambio generates, you’ll notice two distinct slambio-colored areas on the clock. Put another way, here is a histogram of spin axis frequencies for the slambio:

The slightly higher spin axes make the slambio look like a changeup or split out of hand, the lower or much higher spin axes make it look like a curve, and the most frequent ones make it look cutter-like.

Yet the slambio actually has below-average movement standard deviation (both vertically and horizontally), which I calculated by looking at the first 59 pitches — the number of ghost forks thrown so far — for every offering in the dataset in order to negate any impact of sample size. It does, however, have above average spin-axis standard deviation, which is what the charts above are showing.

Why doesn’t this result in more movement variety? Well, as noted on the Baseball Savant spin direction leaderboard, movement is determined both by spin axis and spin rate. Think of spin axis as the direction of the spin and spin rate as the magnitude. The slambio has a below-average spin rate standard deviation. At the same time, spin rate standard deviation didn’t correlate significantly with swinging-strike or called-strike rate. Whether due to its influence on movement or deception, spin axis standard deviation did correlate with higher swinging-strike rates. When looking at range rather than standard deviation, the slambio has an even-wider variety of spin axes. Spanning 272 degrees, its spin axis range ranks 76th in the dataset. The small, separate bumps we saw on the spin axis and direction charts come into play here.

Recall your geometry days now. A one-degree difference can mean two very different things depending on where you are on the unit circle; that can mean going from 360 degrees to 0 degrees at the top, and merely from 180 to 181 degrees at the bottom. In other words, this explains why curveball spin axes can be either very high or very low, and why linear standard deviation doesn’t really make sense for analyzing the variation in spin axis. It wasn’t feasible to factor this into the pitch comparison model because it caused more scaling issues, but I was able to apply it here by calculating the circular standard deviation. In this regard, the slambio had the 83rd-highest deviation in the dataset. Higher circular standard deviations in spin axis correlated with both higher swinging strike and called strike rates.

The ghost fork, on the other hand, made up for what it lacked in spin axis deviation with above-average horizontal and vertical movement standard deviations. Higher movement deviations also correlated with higher swinging-strike and called-strike rates. For all of the important attributes for predicting swinging-strike rate, including velocity deviation and excluding spin rate deviation since it didn’t correlate with anything, I made a composite measure of deviation. The ghost fork rated well, ranking 157th, but the slambio’s spin axis fortitude couldn’t overcome its lack of movement deviation, placing it 656th.

Even though the ghost fork performed well here, for now I would say that for it and the slambio the most important factor for success is uniqueness, which is where the two pitches truly stand out. And though my pitch comparison model still has room to grow in optimally applying spin axes non-linearly, both pitches are equally distinct in the other dimensions I used for the comps. Time will tell whether the pitches will remain successful as the league becomes familiar with them, diminishing the impact of their uniqueness, but if I had to guess, the ghost fork might fare better in the long-term simply because Senga is more adept at manipulating its movement. Hopefully both hurlers can stave off that decline, but in the meantime, let’s enjoy each of their fascinating pitches, as well as Hamilton’s inspiring comeback story.