NBA Best Bets
Player props where our chance of winning beats the bookmaker's price, once its margin is taken out.
A best bet is a price we think is too long. The projection is turned into a probability of clearing the line, the bookmaker's price is turned into a probability once its margin is stripped out, and the two are blended — weighted heavily towards the price, because last season's record showed the market was the better guide. A pick publishes only when that blended chance, multiplied by the price, still returns more than it costs. Every line and both sides are priced, not just the longest odds on offer. Players who have missed a recent game or carry an injury tag are left off, and so are picks where the model disagrees with the market by an implausible margin: last season, those were the heaviest losers.
Frequently asked
Our chance of the pick winning, multiplied by the decimal price, minus one. A 50% chance at 2.10 is worth +5%. Our chance is the market's probability, with the margin removed, moved part of the way towards the model's — so a pick needs both a genuine disagreement and a price long enough to pay for it.
Because that is what the record showed. On last season's playoff picks the bookmaker's price, margin removed, predicted results better than the projection did, and the picks where the two disagreed most lost the most. The model is given a small, deliberate say, and the results page tracks whether it earns more.
It needs posted prices, and it would rather publish nothing than a bet it expects to lose. Player markets go up about a day before tip-off, and on plenty of days no price is long enough to clear the bar.