Show HN: DeepShot – NBA game predictor with 71% accuracy using ML and stats (github.com)

1 points by francescosacco 1y ago ↗ HN

Hey everyone, I’m an NBA fan and Python dev, and I recently built DeepShot — a machine learning model that predicts NBA game outcomes with about 71% accuracy based on historical stats and rolling performance metrics (EWMA). It features: Real NBA data from Basketball Reference Exponentially Weighted Moving Averages to track momentum Interactive NiceGUI interface with team comparison and predictions Full Python stack and open-source (MIT license) Here’s the GitHub repo: https://github.com/saccofrancesco/deepshot And if you like it, here’s my Buy Me a Coffee: buymeacoffee.com/saccofrancesco Would love any feedback — especially from folks who’ve built sports models or worked on real-time stat tools. Also open to ideas on where to take this next (player-level modeling? betting advice dashboard?). Thanks!

1 comment

[ 3.0 ms ] story [ 7.0 ms ] thread

gus_massa 1y ago ↗

It's a small improvement since the previous big post https://news.ycombinator.com/item?id=43685122 (17 points | 44 days ago | 14 comments) Is there any interesting takeaway from the improvement?

I agree with some of the comments in the previous post and it would be nice to compare to easy methods, like just the one with more wins (in the last 5 years?)

Also, a few repost are fine, but with too many repost people will get angry and flag your post and write angry emails to dang/tomhow and they will ban youe user and your site.