Show HN: DeepShot – NBA game predictor with 70% accuracy using ML and stats https://ift.tt/oMQinGf

Show HN: DeepShot—NBA game predictor with 70% accuracy using ML and stats I built DeepShot, a machine learning model that predicts NBA games using rolling statistics, historical performance, and recent momentum—all visualized in a clean, interactive web app. Unlike simple averages or betting odds, DeepShot uses Exponentially Weighted Moving Averages (EWMA) to capture recent form and momentum, highlighting the key statistical differences between teams so you can see why the model favors one side. It’s powered by Python, XGBoost, Pandas, Scikit-learn, and NiceGUI; runs locally on any OS, and relies only on free, public data from Basketball Reference. If you’re into sports analytics, machine learning, or just curious whether an algorithm can outsmart Vegas, check it out and let me know what you think: https://ift.tt/WJmEPQK https://ift.tt/WJmEPQK November 9, 2025 at 01:19AM

Commentary:
What makes DeepShot compelling isn’t just the 70% accuracy claim—it’s the way it reframes NBA prediction as a problem of momentum detection. Most models or betting odds rely on static averages, which flatten out the story of a season. By using Exponentially Weighted Moving Averages (EWMA), DeepShot captures the “hot hand” effect and recent form in a mathematically principled way. That means the model doesn’t just say who is better overall—it shows who is better right now.

The other interesting angle is transparency. Instead of being a black box, DeepShot highlights the statistical differences between teams so users can see why the model leans one way. That’s a big deal in sports analytics, where trust often comes from being able to interrogate the numbers.

Technically, the choice to build it with Python, XGBoost, Pandas, Scikit-learn, and NiceGUI—and to run it locally with only free public data—makes it accessible. Anyone can experiment, tweak, or extend it without worrying about proprietary APIs or hidden datasets. That’s a refreshing contrast to most prediction tools, which are either closed or tied to betting platforms.

The bigger question is can models like this eventually outperform Vegas odds consistently, or are they better suited as decision support tools for fans and analysts? Either way, DeepShot is a fascinating experiment in combining machine learning with sports intuition, and it opens the door to richer, more interactive analytics for basketball.