AI Predictions · NBA
Every pick ships with a measurable edge
Our model compares calibrated win probabilities against de-vigged market odds. When it disagrees with the price, you see it — edge, confidence and stake size. Free tier gets the featured pick; Premium unlocks the full board.
Today's Board
how is this graded? →How the edge is made
1 · De-vig the market
Bookmaker odds contain a margin. We strip it with proportional de-vigging to get the market's true implied probabilities — the honest benchmark.
2 · Model the game
Gradient-boosted trees trained on years of results: team strength, recent form, rest days, back-to-back fatigue — strictly pre-game data, no leakage.
3 · Bet only the gap
When model probability beats the market price by ≥ 4 points, it's flagged. Stakes follow quarter-Kelly sizing, capped at 2% of bankroll per bet.
Straight answers
Does this guarantee I will make money?
No — and any service claiming guaranteed profit is lying to you. This is a probability model: even a genuine 5% edge loses roughly 4 bets out of 10, and losing streaks happen. We publish calibration and ROI (including drawdowns) precisely so you can judge for yourself. Only stake money you can afford to lose.
What do I get on the Free plan?
The featured pick of the day with full edge and stake detail, plus the complete performance dashboard — accuracy, calibration, ROI by edge bucket and every settled result. Premium unlocks the whole board: every pick, every edge, every stake.
How do you prove the track record is real?
Every prediction is logged server-side with a timestamp before kickoff and graded afterwards — wins and losses stay in the record. The performance page is rendered straight from that log via the public API, so it can't be cherry-picked.
Can I cancel anytime?
Yes. Premium is a rolling monthly plan and refunds follow our published refund policy. No dark patterns, no retention hotline.