Minesweeping: Looking for Baseball’s Next Popular Pitch

Last year, the sweeper took baseball by storm. Fast forward to this season, and 4.2% of all offerings through the first half have been sweepers, according to Statcast, nearly twice as many as last season. But I have my issues with MLB’s pitch classification system, and it’s been well documented that under their sweeper umbrella there are multiple versions of the pitch; the Yankees’ staff alone threw several different variations last season. Plus, if the number of different names for the pitch (whirly, rising slider, etc.) is any indication, other teams have their own iterations, too. Qualms with MLB’s system aside, if we want to look for the next sweeper, it’s a given that MLB won’t have a classification for it yet anyways.

Why should we look for the next sweeper? The pitch was extremely effective last year, saving pitchers 0.56 runs per 100 tosses. Even this season, as usage has nearly doubled, the sweeper is still saving pitchers 0.18 runs per 100 tosses. But in order to look for the next one, we first have to ask: what makes all sweepers… sweepers? MLB relies on grip and self-reported pitch identifications for their classifications. In the absence of those, we can use velocity, spin rate, spin axis (in three dimensions), and movement (in two) to identify a new pitch.

Even though some teams might throw multiple versions of a pitch, I still think that our best bet to find a new pitch type is by honing in on individual teams. As with the sweeper and its early adopters, teams that discover an effective new pitch will want to teach it to everyone they can, uniformity of pitching looks be damned. In other words, if a team has multiple different pitchers throwing a specific pitch, they must like it so much that their affinity for it outweighs the cost of having pitchers that don’t contrast (which seems to reduce effectiveness).

So I returned to Euclidean distance, which I used earlier this year to estimate the uniqueness of both Kodai Senga‘s ghost fork and Ian Hamilton’s slambio. In short, the method is a multivariate version of the calculation for the distance between two coordinate (x-y) pairs on a Cartesian plane, the variables in this case being the aforementioned velocity, spin, and movement characteristics, and the actual points each comprising an individual pitch (thrown by a right-hander, for simplicity, and at least 100 times thus far this year).

But whereas before I was comparing the ghost fork and the slambio to every other offering league-wide, this time I only compared each pitch to every other offering team-wide. Due to the smaller samples (there were only 942 qualifying pitches, so about 31.5 per team), I took the mean Euclidean distance of the closest three pitches rather than the closest five as I did previously. The least unique pitches on the team level were all fastballs, which didn’t surprise me since four-seamers are still the most common pitch type, and fastballs in general (i.e., including sinkers) don’t have as wide a movement distribution as breaking and offspeed pitches. Still, certain teams showed up as having particular fastball types. What do I mean by this? The Nationals, Mets, White Sox, and Padres each had at least three of the 20 least unique fastballs (Washington and Chicago each had four):

This graph doesn’t include every pitch characteristic I used, but you can still see where each of these teams’ lack of uniqueness comes from. For the Nats, they have one bunch of riseballs headlined by Josiah Gray‘s (-11, 15), and a group of sinkers headlined by Jake Irvin‘s (-17, 8). The Mets have a cadre of riseballs most concentrated around Jeff Brigham‘s (-9, 18), which is right next to the White Sox cluster around Keynan Middleton‘s (-9, 17). Lastly, the Padres have their own group of purer riseballs (i.e., less horizontal movement), led by Seth Lugo’s (-7, 18).

This isn’t quite what I’m looking for. The Mets probably have the most heavily-concentrated group of fastballs, but that group has mostly average movement, and the Padres have the most interesting cluster in terms of movement, but the rest of their heaters are more far-flung. These aren’t great candidates for the next sweeper.

What if we turn our attention to breaking balls and offspeed pitches? This is where things get a bit more interesting given the wider distribution of pitch movements and the lower frequency of each particular pitch type. The Mets were repeat offenders with three cutters in the top 20 least unique pitches, but they were joined by new arrivals in the Cardinals, Rockies, and Orioles:

If you compare the axis labels for this graph to the last one, you’ll notice that the distribution of horizontal movements is indeed much larger (that of vertical movement is roughly the same). That also means the clusters are generally not as close together, but they’re still there if you know where to look. The Mets cutter cluster can be found around 1, 9, near Senga’s, with the Baltimore cutter cluster not far away around 1, 10 and Kyle Gibson‘s cut-piece. Next is the Colorado cutter-slider thing, which comes in around 4, 6 thanks to Chase Anderson. And finally, another cutter (slash slider thing), this one courtesy of St. Louis. It resides around 5, 2 with Drew VerHagen’s leading the way.

The cutter is in vogue this year; that isn’t news. Cutter usage hasn’t jumped as much as its sweepier counterpart, but it’s also at its highest in the pitch tracking era: 7.8%. That’s probably why more of them show up here than sweepers; there simply aren’t enough of the latter being thrown. But cutters also have a smaller movement distribution than sweepers, so they’re even less likely to be significantly different when they’re taught by the same coaches. Still, if I expanded my reach to the top-23 least unique breaking and offspeed pitches, three Astros sweepers would make the cut.

Despite their tighter movement distribution, there’s clearly variation among cutters as well. The Orioles and Mets throw similar ones with a good amount of rise, but the Cardinal Cut has more horizontal movement and drop, and the Colorado Cut is somewhere in between. The latter two might be better thought of as gyro sliders, but those pitches are also in vogue for the same reason as the cutter: to serve as a bridge between the sweeper and fastball.

Even if there isn’t a new sweeper variation in the works, new cutter offshoots and gyro sliders might have the sweeper to thank for their own uptick in usage. The sweeper has changed how we think about arsenal optimization and how each pitch impacts every other offering. But, before we start labeling gyro sliders and Colorado Cutters, it’s worth noting that there are other ways to identify new pitches. As I pointed out when assessing the least unique fastballs, one could emphasize league-wide in concert with team-wide uniqueness, though that might lead you too far in the direction of pitches that are harder to teach and aren’t likely to become popular.

I chose to focus on the team-specific aspect because I think it indicates true belief in a pitch type (especially if those pitches are concentrated like the Mets’ fastballs), given that it also comes with the risks of having too many of the same pitch; in fact, the uniform-fastball teams have all had difficult seasons, as have three of the four uniform-breaking ball teams. The Orioles are the one team that’s been good, but their pitching hasn’t been. Every team I mentioned except the Padres is in the bottom half of the league in ERA, and even they have vastly underperformed relative to expectations, with the other two most noteworthy underperformers this year — the Mets and the Cardinals — also appearing in my investigation.

Maybe this has also been an exercise in identifying teams who need to diversify the looks they give hitters throughout a game or a series. Adding release point to the Euclidean distance equation would make this even more suitable for that task. But I chose to forego including release point here because pitch types aren’t traditionally defined by it. Perhaps that’s a mistake, but it’s one to address in a future piece.