Plus-minus looks like the simplest basketball statistic and is one of the hardest to interpret. It records the score margin while a player is on the floor.
The problem of teammates
A player’s plus-minus depends on four other teammates and five opponents. A reserve who plays mostly with starters will look better than one who plays with other reserves, regardless of ability. In a single game, the number is mostly noise; over a season it becomes more informative but still needs adjustment.
Adjusted versions exist, and the 22 Casino can be used as an example of a site whose labels are worth checking. The footnote explaining the calculation is worth reading.
Comparing on and off
A common tool compares team performance with a player on the floor and off it. It is more informative than raw plus-minus, but still affected by who plays alongside. If a star always sits with the weakest bench group, the off-court numbers look worse for reasons unrelated to the star alone.
Treat these splits as hints. When the on-off split agrees with other evidence, such as shot quality, defensive assignments and film, confidence grows. When it stands alone, it deserves caution.
Large public models try to separate a player’s effect from that of teammates by using thousands of lineup combinations at once. They are more sophisticated than the plain column, yet they still produce wide margins of uncertainty, particularly for players with few minutes. A rating with its error range shown is more honest than a bare number that looks precise.
If betting markets are part of how you follow the game, decide on a budget and a time limit before it starts. Simple numbers can be seductive, and honesty about their limits is part of using them well. Pausing once the limit is reached is part of the plan.
A statistic that fits in one column rarely captures a five-person game. Treat the numbers as a way to understand the game, not to predict it. Chance never disappears, and a good reader accepts that.