16 seasons of NBA data. Nearly 150,000 graded props.
When the math disagrees with the sportsbooks, we bet.
Join the waitlist to lock in founding member pricing before it opens to everyone. The free card on X isn't going anywhere.
The full card goes out on X with the line, the price and the books it's available at. Some nights it's a handful of plays. Some nights it's none. Late games get their own card before tip-off.
The next morning, every pick is graded here, wins and losses, in the same place.
Every day, every player prop on the slate runs through the same four-stage pipeline, built to find the lines the sportsbooks got wrong.
Machine-learning models trained on every NBA game since 2010 forecast each player's role and production for tonight, factoring in pace, rest, opponent defense and how injuries reshape the rotation.
Each forecast becomes a full probability distribution of the stat line, not a single number, calibrated against years of real outcomes.
Every sportsbook's line is converted to a probability and the vig is removed to find the market's fair price. Then it's our number against theirs, prop by prop, book by book.
Only edges that clear a strict threshold, at prices you can actually bet, make the card. Every pick is tracked against the closing line, and the models are retrained every week.
If a record can't be checked, it doesn't count. Here's exactly how ours is kept.