Stephen Sutton-Brown
@srbrown70.bsky.social
2.8K followers 120 following 1.7K posts
Braves, heat maps, cars, tractors, and music Creator of StuffPro and Arsenal Metrics at Baseball Prospectus He / him
Posts Media Videos Starter Packs
srbrown70.bsky.social
ok it is def having an impact on the model. Here's average distribution of Pitch Type Probabilities when a pitcher throws their FS if we calculate PtP only using each offering's release point distributions. Next is for Bryce specifically when he throws his FS. FS is much more obvious for him.
srbrown70.bsky.social
later tonight i can see how his pitch type probability distributions compare to league if I look just at the release portion. basic idea is what's the probability this pitch is any one of his offerings given the release point from batter's POV plus some additional uncertainty
srbrown70.bsky.social
that would stick out, as the model calculates pitch type probability based on release, on release and initial trajectory, then finally (what's used in production) release, initial traj. and usage rate. it does look a little less apparent from batter POV though
srbrown70.bsky.social
it's like when kids grow faster than their coordination can keep up
srbrown70.bsky.social
i'm sure someone has written about this but strange to me how much bryce uses his four-seam given how broad his arsenal is
srbrown70.bsky.social
it is, i can't get over how insane it is
srbrown70.bsky.social
If you sort instead by lowest expected run value if swung on it looks less impressive since it had a lower probability of a whiff than these others. That Devers homer off Strider was insane
srbrown70.bsky.social
On Judge's HR against Varland, PitchPro estimated an expected wOBACON of just 0.139 if put in play. That's the second lowest E[wOBACON] for any home run in 2025.

Note that Bleday's homer off Matsui came off the bat at just 94.8 mph and snuck over the wall.
srbrown70.bsky.social
incredibly useful
mikepetriello.bsky.social
Another new thing: we add so much cool stuff to Baseball Savant that it's not always easy to keep track of, so here's a good 'whats new' page we'll do our best to keep updated

--> baseballsavant.mlb.com/changelog
Savant Changelog
Baseball Savant
baseballsavant.mlb.com
srbrown70.bsky.social
*logging on to read some very serious baseball writing*
srbrown70.bsky.social
are you able to say which baseball projects you've used BART for?
srbrown70.bsky.social
Repurposing the SaberSem pitch detection algo to see what pitch types and movements a batter might expect when they see Trey Yesavage's SL out of hand (h/t @enosarris.bsky.social on Rates and Barrels)
srbrown70.bsky.social
how i'm spending my time between innings
srbrown70.bsky.social
Pulling the Arsenal Metrics for it, lots of deception
Reposted by Stephen Sutton-Brown
srbrown70.bsky.social
He saw my tweets about Skubal
bsky.app/profile/mich...
srbrown70.bsky.social
I'm generally of the belief that over a season challenging guys in zone is worth the extra hard contact (once again looking at you, Blake Snell): You to prevent walks and falling behind in the count, and you avoid running your pitch count up. But in a single series I'm not sure.
srbrown70.bsky.social
Will try to do a few of these post-mortems this October. I think it helps getting a better sense of what the model sees and how well that matches my own intuition.
srbrown70.bsky.social
I'm generally of the belief that over a season challenging guys in zone is worth the extra hard contact (once again looking at you, Blake Snell): You to prevent walks and falling behind in the count, and you avoid running your pitch count up. But in a single series I'm not sure.
srbrown70.bsky.social
Something to remember about pitch models is they (correctly) don't think a HR is the most likely outcome of a meat ball.

Here are the two HR Skubal gave up thru the eyes of PitchPro. Model expects <50% chance of in play, & sees a 1B or 2B as far more likely than a HR if in play.
srbrown70.bsky.social
it's time to spidey save the day!
Reposted by Stephen Sutton-Brown
srbrown70.bsky.social
i need closer games to justify my not doing yard work
srbrown70.bsky.social
something i've thought about a lot since doing more research is whether the Arsenal Metrics are assuming too little visual uncertainty at release point. Freddy Peralta is a good example of that. He's got distinct arm angles for FA/CU and CH/SW, yet it doesn't appear to be hurting his results.
Reposted by Stephen Sutton-Brown