Algorithmic sports models
The math makes the pick. Not a hunch.
Everything published here is the output of a statistical model — no touts, no gut calls, no takes about who wants it more. Each algorithm is built around a specific, testable idea, measured against seasons of real results, then locked down and run every day against live market prices. When a model has an edge, the numbers show it. When it doesn't, the numbers show that too — and both stay posted.
01
Every algorithm starts with a specific claim about what predicts an outcome — recent form, matchup structure, pace, whatever the sport actually rewards — and turns it into numbers a computer can score.
02
The model is fit on past games, then tested against seasons it never saw during development. If it survives that, its settings are locked. Nothing gets quietly retuned after a bad week — that's how a model fools the person who built it.
03
Each day's picks are logged at the price shown and graded automatically once the games finish. Every model page carries its own running ledger: record, units, return, and whether its confidence has held up.
Basketball models are in the research stage. Same process, same standard of proof — nothing appears here until it has been tested against seasons it wasn't built on.
Football is next in the queue. New algorithms get added as they clear the bar, so the menu above will keep growing through the year.
What these numbers are. Results across this site are tracked in paper units at the prices recorded when each pick was published. These models are experiments, not advice — they're posted so their performance can be judged honestly, wins and losses alike. For entertainment purposes; bet responsibly.