Skip to content

NBA · 2026-27 · Thu, May 21

StatGrades AI Model

One model, a call against the spread on every NBA game with a line: team ratings from every game's closing line and result, plus home court, against the books' lines. Graded against the final score. In testing it didn't beat the closing line, so read every pick as a lean.

These picks are for information and entertainment, not financial advice, and no pick is a sure thing. StatGrades doesn't take bets or link you to a sportsbook account. Odds are shown for information and can change at any moment; check with the book before you bet. 21+ only where legal. Gambling problem? Call 1-800-GAMBLER.

The record is open. Every pick is free with All-Access this season.

All-Access shows every pick in full: the side, the model's line and the predicted score for every game. It's $12.99 a month, and free with an account for the rest of the season.

No picks for this night yet: picks are made once the lines are posted, usually the day before.

StatGrades AI Model record

Every NBA game with a line: each team rated from this season's closing lines and results (and, early on, last season's), plus home court, compared with the market. Re-run hourly until tip-off, then graded at its closing line against the final score. Tested before launch on 2025-26, a season it never saw: every lean went 528-538 (49.5%) against the closing line, its confidence 3+ calls 168-198 (45.9%). That's no better than a coin flip: the NBA market already knows what these ratings know. Every pick is a lean, with no top calls and no over-unders, and this live record is published either way.

Top calls

None

not made: failed the test

Against the spread

10-7

59% of decided picks

Over-under

None

not made: failed the test

Straight up

14-3

82% of decided picks

Against the closing number

Whether the market moved toward a pick after it was made, measured against the number the game actually closed at. A record takes a long time to mean anything; this says something after a few weeks. Positive is good.

Spread

−0.12 pts

market moved our way on 1 of 17

The StatGrades AI Model rates every team in points against an average NBA team on a neutral floor, fitted to every game's closing line and final margin this season (a result never counts as more than 20 points better or worse than its line), with older games counting less and last season fading out over the first weeks. It adds about 2.5 points for home court and takes the side of the gap between its line and the market's. On each card the Gap is that difference in points against the books' live number. It knows nothing about injuries, rest or back-to-backs, which the market prices in, and in testing its gaps didn't beat the closing line, so a big gap is a measurement, not a promise.

The boxes on each game: the Predicted Score is the model's projection, with an arrow at the team it has winning. Before tip-off each box is the best line on offer, with the book that has it, and the percentage is our chance at that line. ML is the model's own chance of each team winning outright; Spread and Total put its projection against the line, with misses the size NBA games usually land from the closing line (about 14 points). Our pick is outlined green, and turns red if it loses. After tip-off the boxes show the lines the picks are graded at.