Twitxy Saved

Alex Caravan

@Alex_Caravan

MLB Data Science Manager @ Penn Interactive | ex-@DrivelineBB | (Small sample ⟦size⟧)
Joined January 2018
1.7K Following 3K Followers 3.4K Posts

Latest posts 5

Alex Caravan @Alex_Caravan
💯💯💯 You realize this when you hop over from player development to pure projections…obviously the two fields are intertwined but if for example you want to build the best possible K% predictor at a league wide level you’re not using Stuff+ or just Stuff+…
Chris Langin Verified account @LanginTots13 Oct 10, 2025
A few random thoughts on Stuff+:

Stuff+ is highly reliable, which makes it predictive in low-information environments (small samples, new arsenals, showcases).

That doesn’t make it universally predictive. Once you have real outcomes and stable peripherals, those explain future performance better than a static Stuff+ value.

Often, weighting each of them can help as you are in the middle ground between minimal information (all stuff+) and say strikeout-walk % stabilizing (projection model).
View quoted post on X
1 replies 1 reposts 23 likes 10K views View on X