Daily NBA picks driven by a six-model ensemble: pace-adjusted efficiency, rest and travel context, lineup-based projections, and live market scanning. Verified track record. No "feel" picks.
Every NBA play on this site comes from a single, repeatable pipeline: a six-model ensemble trained on three-plus years of game-level NBA data, real-time sportsbook odds, and a fixed set of edge thresholds that decide whether a game becomes a play.
The ensemble produces a fair-value win probability and projected margin for every game on the slate. The model then compares that probability to the implied probability priced by the market across moneyline, spread, and total. When the gap clears the threshold and the rest, travel, and lineup context confirm — the game becomes an official play with explicit unit sizing.
Picks are not posted until starting lineups are announced. NBA outcomes turn on availability, and we don't guess.
NBA edges live in the gap between what the market prices at tip and what the underlying efficiency, fatigue, and lineup data actually project. The ensemble weights four core inputs:
Raw points-per-game is a noisy stat in a league where pace varies widely between teams and across games. The model uses offensive and defensive efficiency adjusted for opponent strength and possessions, then projects expected points per 100 possessions for both sides at the projected pace of the matchup.
Back-to-backs, three-in-fours, time-zone changes, and altitude all materially shift NBA outcomes. The model encodes rest-day differentials and recent travel mileage into the projection — a team on the second leg of a back-to-back after a coast-to-coast flight is not the same team that played at home two nights ago.
Stars sit. Rotations change. The model recalculates probability the moment confirmed lineups drop, weighting each player's on-court impact and the lineup's net rating. A star-rest situation that the market hasn't fully priced is one of the cleanest NBA edges.
The model only fires when its projection beats the closing-line implied probability by enough to clear the vig and absorb variance. The bet is the gap, not the favorite. A play on a moneyline dog at +180 with a 42% projected win probability is mathematically more valuable than a -250 chalk priced exactly at the model's projection.
Each NBA play arrives with a fixed structure so you know exactly what you're betting and why:
Plays are released after starting lineups are confirmed by both teams — typically 30 to 60 minutes before tip. We do not publish blind picks before lineup confirmation; star availability is too large a factor in NBA outcomes to project around.
Coverage runs full regular season plus playoffs. Slate volume varies. A typical NBA night with 6 to 8 games might produce two to four plays. A heavy lineup-news evening can fire more. Some nights the model stays out entirely — the system is calibrated to find real edges, not to fill a daily quota.