Once my model all goes to shit by week 2 I might code this logic into it insteadDublin_Flyer wrote: ↑Sat Aug 09, 2025 12:11 amMore advanced than me Tuple, I just have a few beers and pick random results I'd like to see.TupleVision wrote: ↑Fri Aug 08, 2025 11:04 pmRolling into the season with a little numbers-powered cheat code
I’m pulling team data from Understat and the FPL API (fixture difficulty + who’s likely to start), blending last season’s baseline with this season’s form, giving a tiny home/favourite nudge, then running a bivariate Poisson to spit out the most likely exact score for each fixture. Paste matches in, predictions out. Not wizardr! just a tidy, squad aware model to keep things smarter than “1–0 every week![]()
So 0 out of 6 for me the first week.
I'm always near the top....of the bell curve![]()
