The short version — the record does the talking
How this works
What we publish
Picks on NBA player stats — points, rebounds, assists, threes, PRA. Each one goes up before tip-off with its line, its odds and the model's probability, stamped with the exact time of publication. After that it is never edited and never removed.
Where picks come from
An in-house statistical model, tested against four full seasons of past games before it ever published a pick. When its number disagrees with the market's by enough, that disagreement becomes a pick. The details stay in-house — the results don't.
How picks are graded
From the official box score, automatically: win, loss, push when the stat lands exactly on a whole-number line, void when the player doesn't play. Flat one-unit staking at the recorded odds. Every result stays on the record, red or green.
What the numbers mean
Units and ROI are computed from the record alone. CLV (closing line value) compares each pick's price with the final price before tip-off — consistently beating the close is the earliest real sign a process works, long before win-loss streaks mean anything.
The honest part
Small samples prove nothing, losing runs are normal, bookmakers limit winners, and the price you get may differ from the recorded one. Some days there are no picks at all — that's discipline, not failure. If the record ever shows the edge isn't real, that gets published with the same prominence as everything else. Never bet money you can't afford to lose.