Hi Product Hunt! 👋 I'm Ali, and I built MatchMind after getting frustrated with football prediction sites that make big claims and never publish their accuracy. The core idea: if you call something a "probability", it should actually behave like one. That means publishing the Brier score (including bad weeks), showing reliability diagrams, and having a post-match audit page that doesn't let us hide the misses. A few things that might interest the data folks here: → End-of-season accuracy drops from ~41% to ~26% due to squad rotation and motivation effects. We added a season_completion_pct feature to compensate. → Draws are the hardest outcome to predict. Our model underestimates them. We publish this openly. → No raw model probabilities ever reach the API — everything goes through isotonic calibration first. Full methodology on the blog: matchmind.football/blog Free during beta. Happy to answer any questions about calibration or the prediction pipeline! 🙏

MatchMind

Calibrated football match probabilities — no tips, ever.

Calibrated football match probabilities — no tips, ever.