Questions, answered straight.

AlphaLine is a statistical projection product for NBA player props. It is not a sportsbook and you do not bet here. The model projects a stat line for each player, compares it to the book's number, and posts the props where it sees an edge. You place any wagers yourself, at whatever licensed sportsbook you use.

A walk-forward Ridge regression projects the stat line and a LightGBM classifier ranks the board. It is trained on past games and scored on the unseen future, so there is no data leakage. It is not a tout, not a person's hand-picks, and not a gut-feel service. The picks come from the model.

More detail on the full pipeline is on the How It Works page.

NBA player props only. That includes points, rebounds, assists, threes, and steals, plus combo markets like PRA (points plus rebounds plus assists) and P+R (points plus rebounds).

Injury and rotation data are scraped every 15 minutes and are trade-aware, so a player who has changed teams is mapped to the correct roster. Stale injury data is one of the fastest ways to get a prop wrong, so it stays fresh right up to the slate.

On the held-out 2025-26 season (the model was never trained on it), it went 231-121, 66.0% win rate, +20.2% ROI at best-of-book price, across 352 picks over 129 slates. That 66.0% is from a season the model never saw. It is the most honest number we have.

Live results are tracked publicly from launch. We publish all of it, wins and losses, at the Track Record. Judge it there.

The 2025-26 evaluation ran the model on a season it was never trained on. That produced 66.0% and +20.2% ROI. It is a real held-out result, not a backtest on training data, but it is still historical.

Live results are real picks posted before tip-off and graded the next morning. That is the ongoing test. Every graded live pick is public, and losses are never deleted.

SNIPER is the highest-conviction tier: fewer plays, tightest edge filter. CORE is the everyday value board: more volume, still edge-ranked. Both are posted before tip-off and graded publicly at the Track Record.

No. No honest model can. Sports outcomes are variable, sportsbooks adjust their lines, and even a genuine edge loses plenty of individual nights. Anyone guaranteeing profit on props is selling you something. AlphaLine gives you edge-ranked projections and a fully public track record; the rest is variance and your own bankroll decisions.

Please read Responsible Play before risking money.

About 2-3 per night on average. The board is capped deliberately. The model ranks the entire slate by edge and only the highest-edge plays make the cut, so you get a short, disciplined list instead of a flood of marginal lines.

Before tip-off, and every pick is timestamped. That timestamp is what makes the track record meaningful: a pick is locked before the game starts, then graded the next morning against what actually happened.

Yes. You can verify AlphaLine before paying anything. The Discord is free to join and the full Track Record is public, so you can watch the board grade itself for as long as you want before deciding to subscribe.

Join at discord.gg/ZggXCsDDRd.

Current plans and prices are listed on the Pricing section of the homepage. Start there to subscribe.

You can cancel anytime from your subscription settings, and you keep access through the end of the period you have already paid for. If you have any trouble, reach us in the Discord and we will sort it out.

The NBA season runs roughly October to June, so the live board is quiet over the summer when there are no games to project. The backtest board and the full track record stay available year-round, and the model is maintained and improved during the break.

The Discord is the best place for questions, support, and anything that is not covered here. Join at discord.gg/ZggXCsDDRd.