You remember the hits, of course. You remember when a pitching model that attempts to take the noise out by focusing on stuff and locations alone helps you find players like Shane McClanahan, Dylan Cease, Sandy Alcantara, Spencer Strider, Zac Gallen, Adam Wainwright, and Drew Rasmussen in the preseason. They’ve been huge this season.

But Pitching+, which is the combination of Stuff+ and Location+ in a third model, is most effective at being fast, so it’s maybe even more important as an in-season tool. It’s predictive quicker than strikeouts minus walks, the other most effective stat in short samples.

This graph says that there’s less error (RMSE) in using Pitching+ to predict rest of season runs allowed than in using strikeouts minus walks (K-BB%), at least until 150 plate appearances. That means if you’re looking at a pitcher who has thrown fewer than eight starts, it’s better to look at his Pitching+ than anything else. That helped us identify pitchers like Cristian Javier, Kyle Wright, Blake Snell, Nick Lodolo, Lance Lynn, Braxton Garrett, Mitch Keller, Austin Voth, Tyler Wells, Tylor Megill, Jeffrey Springs, and Alex Cobb this season. The model also helped identify struggling starters that weren’t necessarily buy-lows, like Lucas Giolito, Josiah Gray, Ian Anderson, Daniel Lynch, Dane Dunning, Kyle Gibson, and Sean Manaea.

But no model is perfect. No projection, no stat, no prognosticator gets it all right. The challenge when looking through the misses, as we will here, is to learn the just right amount. Sometimes there are just outliers bearing little in terms of new information, and sometimes the misses mean an adjustment must be made.

 

One pitcher we were definitely too high on

José Berríos, Blue Jays (5.37 ERA / 102.9 Stuff+ / 103.8 Location+ / 105.5 Pitching+)

To be fair to Pitching+, Berríos’ Stuff+ is down this year (from 105.9) and so is his location (from 106.8), so it did detect a downturn this year. But Berríos went from a 3.52 ERA to a 5.27 ERA and all of his ERA estimators jumped with it. He’s earned this poor year by striking out fewer batters and giving up the most homers of his career. It’s not about his home park, as he’s given up 18 homers away from home and 11 at home, and it’s not even about his division, as he’s given up three homers in three starts in New York, and that’s the homer-friendliest park in the AL East.

Sometimes we see these choices as black and white, it’s either luck, or the model just missed on him, but the truth could be in between. His four-seam fastball locations are slightly worse this year, and his four-seam has generally been hit a lot harder, and the team around him seems to think it’s about four-seam command. One-season home run rates are some of the biggest sources of noise in the business, so maybe it’s just one bad year. There’s also something to be said for a guy who has one secondary pitch with above-average Stuff+ losing any command or stuff on his fastball. The interaction between the primary pitch and the rest of the arsenal is definitely something worth studying.