All scores →
Loading…
Loading…
Loading…

Plus-minus, on/off and why the raw number lies

By 4Basketball Staff · · 11 min read

Plus-minus records how a team fares while a player is on the floor. Learn how it, on/off splits and adjusted versions work, and why the raw figure often says more about teammates and luck than the player.

Of all the numbers in a basketball box score, plus-minus has a rare appeal. It does not ask you to weigh rebounds against assists or decide how much a missed shot costs. It simply records what happened on the scoreboard while a player was on the court. If a player's team outscored the opponent by eight points during his minutes, his plus-minus is plus 8. If the team was outscored by five, it is minus 5. The statistic claims no theory and requires no assumptions, which makes it feel like the one number that cannot lie.

It can, and it often does. The trouble is that the scoreboard reflects ten players, not one, and the player whose name is attached to the figure had no say in who his teammates were, who his opponents were, or how the shots happened to fall. A bench player who shares the floor with four starters can post a superb plus-minus without doing much. A star who plays all night beside weak teammates can post a poor one while carrying the load.

This article explains what plus-minus is, how on/off splits refine it, why the raw numbers mislead, how analysts have tried to correct the problems, and how a careful fan can still extract value from the figure. It is a good example of a general lesson in basketball analysis: a number can be perfectly accurate as a description and still be a poor measure of what you actually want to know.

The basic statistic

Plus-minus is the net change in the score while a player is in the game. It is added up across a game, a season or a career. Its attraction is that it captures every effect that matters, including those that the box score misses: a defensive stop that creates no steal, a screen that frees a shooter, a pass that leads to a pass that leads to a basket. If a player's presence tends to coincide with the team doing better, the plus-minus will tend to be positive.

The idea was borrowed from hockey, where the statistic has a long history, and was adapted to basketball. For many years it was not part of the standard box score, but it became widely available once play-by-play data allowed it to be computed for every player in every game. It now appears in game summaries across the sport, and it is a standard element of lineup data.

The raw figure has one structural feature to keep in mind. Because five players are on the floor for each team, a single basket moves the plus-minus of ten players at once. A team that outscores its opponent by 20 in a game will have a collection of positive numbers, and the star who played 38 minutes and the reserve who played 12 in a stretch when the starters dominated will both be positive. The number shows a team's result, divided among those present.

On/off: adding a comparison

The most useful refinement is the on/off split. Instead of asking how the team did while a player was on the floor, it asks how the team did with him on the floor compared with how it did with him off. A team's net rating with a player on, minus its net rating with him off, is the on/off figure.

The advantage is that the comparison gives a baseline. A raw plus-minus of plus 8 is hard to interpret. A team that is plus 8 per 100 possessions with a player on the floor and plus 2 with him off is a team that gains 6 points per 100 possessions from his presence. That is a more informative statement, and it is what analysts mean when they speak of a player's on/off impact.

Net rating, the per-100-possession version of point differential, is used for this so that differences in the amount of playing time and in tempo do not distort the comparison. You can read about how the underlying ratings are computed in the article on offensive and defensive rating, and about their summary in the article on net rating.

On/off data is also separated into offense and defense. A player may have a large positive effect on the team's offense and a negative or neutral effect on its defense, or the reverse, and the split makes that visible. This is one of the few places where a statistic gives a window on defense without relying on blocks and steals.

Some famous cases show why on/off analysis attracts interest. Manu Ginobili's Spurs teams were often reported to be markedly better with him on the floor than off, a pattern that fans and analysts noticed even when his box-score numbers looked modest as a reserve; his career is described in the article on Manu Ginobili. Steve Nash's Phoenix offenses were widely observed to be far less potent when he sat, a pattern that fit his reputation as an offensive engine, discussed in the article on Steve Nash. In both cases the data appeared to confirm what viewers believed they were seeing: a player whose impact exceeded what traditional statistics showed.

Why the raw number lies

Now the problems, which are substantial.

Teammates

The biggest problem is that a player's plus-minus depends heavily on the four people he plays with. A reserve who is regularly paired with the best starters will tend to show a strong figure, since those units are strong. A starter who is regularly paired with weaker reserves will show a weaker one.

Consider a coach who subs the best player out for a few minutes in each half, and during those minutes the team plays with its other starters and a couple of reserves. The player's off-court numbers reflect these odd units, which might be better or worse than the team's typical lineup. If the team is deep, the off-court numbers look good and the star's on/off looks small. If the team is thin, the off-court numbers look poor and the star's on/off looks huge. The difference is a feature of the roster rather than of the star.

Opponents

The same goes for opponents. A player who plays his minutes against the opponent's starters is facing tougher competition than one who plays against reserves. A starter's raw plus-minus is therefore penalized relative to a reserve's, even if they are equally good.

Sample size and noise

A subtler problem is that a plus-minus figure from a single game is mostly noise. Over 30 minutes a player might see a few dozen possessions, and the margin in that span can swing on a few lucky or unlucky shots. A player can be plus 15 because the opponent missed a series of open threes while he was on the floor. A season has many more possessions, but even a season's worth of on/off data has a considerable noise component, especially for the "off" part, which may be built on a small number of minutes. The article on small samples covers why this matters.

The three-point shot makes this worse. Because a team's opponents' three-point percentage varies from stretch to stretch in ways that are partly out of anyone's control, a lineup's defensive numbers can swing for reasons unconnected with the people on the floor.

Garbage time and context

Minutes at the end of a decided game are played differently from the rest. Reserves who enter then often rack up plus-minus figures that reflect a game already decided. A starter who rests in the fourth quarter of a blowout can see his number frozen at an excellent value, while a reserve who plays the final minutes is as likely to lose ground as gain it.

Role

The star who is asked to carry a poor team, playing against opponents' best players, can post a negative plus-minus and still be doing exceptional work. The figure tells you how the scoreboard moved, not whether the player was doing more or less than anyone else could have.

An example in the abstract makes the point. Imagine two players, one on a dominant team and one on a poor team. The first has a plus-minus of plus 6 per game and the second minus 5. Does that mean the first is better? Not necessarily. The first may be a decent role player on a team that would be plus 6 whoever filled the spot. The second may be the best player on a team that would be minus 15 without him. The raw numbers are accurate and misleading together.

The attempts to fix it

Analysts have worked for two decades on ways to extract a cleaner signal.

Adjusted plus-minus. The idea is to use all the lineup data from a season and estimate each player's contribution while accounting for who else was on the floor and who the opponents were. In effect, it solves a very large set of equations, one for each stretch of play with a fixed set of ten players, to find the value for each player that best explains the results. Researchers including Dan Rosenbaum developed early versions in the 2000s. The approach is conceptually appealing because it untangles teammates and opponents. In practice, though, it needs a great deal of data, and the estimates for individual players come with large margins of error.

Regularized adjusted plus-minus. To deal with the instability of the earlier method, later versions add a statistical technique that shrinks extreme estimates toward the average, on the logic that most players are not far from average and that a large estimate based on limited data is probably inflated. This improves reliability considerably and is the basis for several public metrics. The shrinkage means that these estimates are more conservative for players with fewer minutes.

Hybrid measures. Others blend plus-minus-type information with box-score data. Box plus-minus, covered in the article on win shares, BPM and VORP, uses the box score to approximate what the adjusted plus-minus would say. That sacrifices some of the plus-minus's ability to capture the unrecorded, but it gains stability.

Lineup-level analysis. Rather than isolating individuals, analysts can look at how particular combinations perform together. The article on lineup data and five-man units explains the potential and the pitfalls, which largely come down to the small number of minutes any given five-man unit plays.

None of these methods is perfect, and the more sophisticated they are, the more they depend on assumptions. They represent a trade-off between the transparency of the raw figure and the bias reduction of the adjusted one.

How to read plus-minus sensibly

A careful fan can still use the figures if the habits are right.

  1. Prefer net rating on/off to raw plus-minus. The per-100 version removes the effect of playing time and tempo.
  2. Look at both the on and the off. An impressive on number with an equally impressive off number tells you the team is good, not that the player is. The difference is what matters.
  3. Ask who the teammates and opponents were. If a player's minutes were mostly with the starting unit against reserves, discount the number.
  4. Check the minutes. The more minutes on and off, the more reliable the figure. Treat anything based on a few hundred possessions as provisional.
  5. Look for consistency across seasons. A player whose on/off is strongly positive in several consecutive seasons, on different rosters, is more convincing than one with a single standout year. The article on regression to the mean explains why an extreme season is likely to be followed by a less extreme one.
  6. Separate offense and defense. If the effect comes from only one end, ask whether it fits what you see.
  7. Compare with other evidence. If on/off disagrees sharply with the box score and with your own viewing, consider why. The disagreement is sometimes real, as with a player who has influence the box score cannot see, and sometimes a product of noise or roster effects.

What plus-minus is good for

Having listed the problems, it is fair to say what the statistic is for.

It is an excellent description. When a coach decides which lineups to use, plus-minus data on those lineups is a legitimate input, with caution about sample size. It is a useful flag: a player with a surprisingly strong or weak on/off figure deserves a second look, either to find the hidden contribution or to find the confounding factor. It is a useful check on the box score: if a player's traditional numbers are modest but his teams are consistently better with him in the game, something is going on that the counting stats miss.

It is also a corrective to the assumption that box-score production captures everything. Sports analytics has long contained the idea that a player's value can lie in things that do not appear in a column, and plus-minus was one of the first attempts to find it. That idea is part of the broader discussion in the article on analytics myths, which warns against both over- and under-reading the numbers.

For anyone who follows a team in the standings and watches how its rotation changes over a season, the on/off figures can help explain why a team is better or worse in certain stretches. They should be treated as clues rather than conclusions.

The deeper lesson

Plus-minus is a nice illustration of a rule worth keeping in mind whenever you encounter a statistic: the question is not only whether the number is accurate but what it is accurate about. The scoreboard figure is exact. What it describes, however, is the outcome of a team activity involving ten players, shaped by lineup choices, opponent quality and a good deal of chance. Attributing all of it to one player is a leap, and the leap grows larger as the sample shrinks.

The history of the statistic is the history of trying to make that leap smaller. On/off splits took the first step by adding a baseline. Adjusted versions took the second by controlling for the company a player keeps. Each step reduced the bias and increased the complexity, and none removed the uncertainty. Good analysis in basketball, as elsewhere, mostly consists of knowing how much confidence a number deserves.

What to take away

Plus-minus records how the score changes while a player is on the floor. It is an honest count of something real, and it can capture contributions that the box score misses. But it is shaped by teammates, opponents, minutes and luck, so the raw figure says as much about the context as about the player.

Use on/off net rating rather than the raw number, check the sample, look at both ends of the floor, and compare with other evidence. When a player's plus-minus tells a story that nothing else supports, be skeptical; when it confirms what the film and the other statistics show, you have found something solid. In either case, remember that the number lies only when it is asked to do more than it can.

Search more on this

Keep reading

Defensive metrics: why defense is hard to measure

Offense leaves a trail of makes and attempts; good defense often leaves nothing at all. This guide explains what defensive statistics capture, what they miss, and how to judge defenders wisely.

4Basketball Staff · Aug 12, 2026 · 11 min read

The three-point revolution in the numbers

The three-point shot went from a novelty to two of every five attempts. This article follows the numbers, the simple arithmetic that drove the change, what happened to accuracy and what it did to spacing, rebounding and scoring.

4Basketball Staff · Aug 28, 2026 · 11 min read

Home-court advantage in the numbers

Home teams win about three in five games, but the number hides a lot. This article converts the edge between wins and points, shows how noisy a single team's home record is, and works out what it is worth in a playoff series.

4Basketball Staff · Sep 11, 2026 · 10 min read