A star sitting the entire fourth quarter of a thirty-point game is the most common way a well-reasoned points prop loses. He was on pace comfortably, the game stopped being a contest, and the coach protected him. Nothing about the matchup read was wrong.

This affects starters on good teams disproportionately — precisely the players whose season averages look most reliable. When you see an otherwise consistent scorer with a handful of unexplained misses in the log, check the final margins on those nights before concluding his form dipped.

The NBA schedule includes back-to-back games, and teams increasingly rest veterans on the second night or cap their minutes. A player who is available but on a restriction is a different proposition from the one the hit rate describes, and the restriction is often announced only hours before tip-off.

Our numbers are computed from completed box scores and today's posted line. They do not know about tonight's minutes restriction. This is the single largest gap between what a basketball hit rate says and what will actually happen, and it is worth checking a beat writer before acting on any line where rest is plausible.

Points are the headline market and the most sensitive to game state, for the blowout reason above.

Rebounds depend on opportunity as much as ability. A game where both teams shoot well produces fewer missed shots to rebound, so a centre can play his usual game and finish short through nobody's fault. Pace and shooting efficiency move this line in ways that are invisible if you only look at the player.

Assists are the most teammate-dependent market in basketball. An assist requires someone else to make the shot. A playmaker can create exactly the same looks he always does and miss an assists line because his team shot poorly from three that night — his passing was identical, the ball just did not go in.

PRA (points + rebounds + assists) combines all three into one number, and PR, PA and RA do the same for pairs. These are generally steadier than any single component, which is the point of them: a quiet scoring night is often offset by rebounds or playmaking, because a player who is on the floor tends to contribute somewhere.

What the combination hides is which part is doing the work. A PRA line that clears consistently might be riding entirely on rebounds from a player whose scoring has fallen off a cliff. If you are using PRA as a proxy for "is this player playing well", open the game log and look at the split — the combined number can stay flat while the shape of it changes completely.

The most volatile market on the NBA board, and it is worth being clear about why: attempt volume is quite predictable, and conversion is not.

A shooter who takes nine threes a game will keep taking nine. Whether four go in or one does is close to a coin-weighting exercise on any single night, even for elite shooters. Hit rates on this market bounce around far more than points or rebounds, and a hot or cold streak in the log is usually variance rather than a change in form.

These sit at 0.5 or 1.5 because the counts are so low, which means a single play decides the outcome. A hit rate on a 0.5 steals line is really asking "does this player usually record at least one steal", and the answer flips on a deflection that could have gone either way.

Low-count markets like these are where small samples mislead most. Treat a strong-looking steals hit rate with more caution than the same figure on a points line.

Because games come thick and fast, the 5, 10 and 14-game windows on an NBA player usually cover a span of two to four weeks and tend to agree with each other. When they do not, the cause is generally identifiable: a rotation change, a return from injury, a trade, or a teammate's absence redistributing usage.

That makes the window comparison genuinely diagnostic in basketball in a way it is not in, say, baseball, where the same disagreement is often just noise.

Every figure on the NBA slate is computed the same way, and the full explanation — what a hit rate counts, how the window changes the answer, and what the numbers cannot tell you — is on our methodology page. If a market term above is unfamiliar, the glossary defines every one we cover.

Nothing here is betting advice. We are not a sportsbook, we take no wagers, and a historical rate is a description of the past rather than a prediction. If gambling stops being fun, help is free and confidential: call or text 1-800-GAMBLER (1-800-426-2537) in the US.