Cutters and Command

OpenCommand, introduced by Doyoung-Tom Kim, is an absolutely incredible open-source dataset that offers a new way to track pitch locations based on target-miss distance, how far a pitch landed from where the pitcher was actually trying to put it, estimated from the catcher’s setup and camera tracking. Without going too much into detail about the workings of OpenCommand, I would highly recommend checking out his work and the great lengths he went to account for camera angles, plate depth, etc.

There are many ways we can go with how comprehensive this data is, and I will continue to dig into the findings. Initially, I wanted to investigate more macro questions about certain pitch types, to see what, if any, trends we can use for fantasy purposes. How much does command correlate to pitch outcomes on specific pitch types? Or does the quality of the pitch, velocity, stuff, movement carry more importance than location, and which pitches relied more on location? Of course, command is an essential quality for any pitcher to be successful, but with metrics like Location+, I have always found that, with the ability to locate a pitch without movement or velocity, even perfectly located pitches can be punished, and for the most part, stuff definitely provides the higher ceiling.

But of course each pitch type’s relationship with stuff, location, and command is different. Four-seam fastballs and sinkers that target the zone, of course, need to be located well, but they are very reliant on stuff in order to miss bats or induce weak contact, while still being in the zone. Big breaking balls will have bigger misses in location, since it is much more difficult for pitchers to know how much that pitch will spin; as we have seen with a pitch like Nolan McLean’s curveball or sweeper, too much spin and horizontal movement can be hard to throw successfully with consistency. So I wanted to test location on OpenCommand with Stuff+ to see the relationship between the two and the pitch outcome.

Location, Location, Location

In this chart, each bubble is one pitch type, with the size of the bubble signifying the sample. Left to right shows the command coefficient; further right means a bigger miss costs you more. The value shown is how much a pitch’s results change for every extra inch it misses its target, once the pitch’s own quality is already accounted for. Up and down shows how much we can trust that number is real rather than a coincidence from the group of pitchers we happened to look at. This statistical confidence comes from that same coefficient: how big it is relative to how much it could plausibly bounce around given the size of the sample. A coefficient built on more pitchers, or one that’s large relative to the scatter in the data, earns more trust; a coefficient built on a small group or a noisy spread earns less, even if the number itself looks similar.

So cutters, perhaps unsurprisingly, are the pitch that is most reliant on being located well. It is the pitch that is least able to lean on stuff, and is the pitch type that has the lowest average Stuff+ score. With velocity generally below the four-seam average, and without the horizontal movement of a slider, sweeper, or even a sinker or vertical movement, it doesn’t have much to fall back on if not located well. If those pitches miss their spot, they are still able to fool bats through other means, whereas the cutter’s success is much more tightly correlated to command.

For the cutter specifically, that number works out to roughly 19 points of xwOBA against for every extra inch missed. The p-value, which drives the y-axis, asks how likely it is that random chance alone would produce a coefficient this far from zero. P stands for probability, so for the cutter, that probability of randomness is about 1%. The four-seamer’s p-value comes in at about 96%, so there is little evidence to show location is what’s driving fastball results; it’s more likely dependent on velocity or movement, whether that be vertical or horizontal, rather than precise location. That same pattern held across every pitch type besides the cutter, meaning we can’t say location reliably costs much on those pitches, especially relative to the cutter.

These are the ten pitchers that the OpenCommand model shows locate their cutter best; command in inches is the average miss distance from their spot. I have also added the pitchers’ cutter xwOBA and Stuff+. Looking at the two tables at either end of the location spectrum, there are a couple of interesting comparisons. Yamamoto and Gore have very similar Stuff+ scores on their cutter, but Yamamoto’s ability to locate that pitch at an elite level results in significantly better results. To take this a level further, McLean’s cutter has a much better stuff grade than Scott’s, for example, but we can see command wins out once again, and Scott is able to throw it much more successfully.

Taj Bradley is the outlier where really elite stuff does eventually win out, but given that of this group of 20 only two were able to score above a 100 Stuff+ score, he is an anomaly in the same way Misiorowski’s cutter is an outlier.

With the fantasy season in its final stages, this isn’t really setting up any potential pickups, but rather something to log for the offseason, especially when pitchers inevitably look to add cutters to their arsenal. Unless their stuff is that of Misiorowski or Bradley, they will need to get a good feel for how to locate their cutter quickly. Logan Gilbert is an example that comes to mind of a pitcher this season who looked to add or reintroduce a cutter to his arsenal in 2026 but quickly abandoned it, as he was not able to utilize it effectively. On the flip side, any lineup facing a pitcher who leans on his cutter but is struggling to locate it is worth targeting.

Being able to get to your spots consistently with any pitch is important, but for those who rely on cutters more heavily, the margin for error is a lot smaller.