Net rating is a team's scoring margin per 100 possessions. Take its offensive rating (points scored per 100 possessions), subtract its defensive rating (points allowed per 100), and that's the number. The 2025-26 Oklahoma City Thunder scored 118.9 and allowed 107.7, a net rating of +11.2 that led the NBA, per Basketball-Reference.
Plus-minus is the same idea for one player, counted in raw points: his team's score minus the other team's while he's on the floor. Offensive rating, the first half of net rating, means one thing for a team and two different things for a player, depending on the site you read.
Net rating counts points per 100 possessions
A possession is one trip with the ball, and it lasts until the other team gets it. An offensive rebound doesn't start a new one. NBA.com's stats glossary says it "simply makes the existing possession longer." Since the ball changes hands after every trip, both teams finish a game with nearly the same number of possessions. That makes points per possession a fair way to compare a team's offense with its defense.
The 100 only keeps the numbers readable. The 2025-26 leaders on Basketball-Reference show the scale:
| Number | What it counts | Example |
|---|---|---|
| Offensive rating | Points scored per 100 possessions | Denver led at 122.6 |
| Defensive rating | Points allowed per 100 possessions (lower is better) | Oklahoma City led at 107.7 |
| Net rating | Offensive rating minus defensive rating | Oklahoma City led at +11.2 |
| Plus-minus | A team's scoring margin while one player is on the floor | Devin Booker scored 70 in a 2017 game and finished −6 |
How offensive rating is calculated
For a team, it's points divided by possessions, times 100. Score 112 points on 100 possessions and your offensive rating is 112; score them on 95 possessions and it's 117.9. Defensive rating runs the same math on the points you allow.
The tricky part is the possession count, since a basic box score doesn't list it. Basketball-Reference's glossary estimates it: shot attempts plus turnovers plus a share of free throw attempts, minus an estimate of the misses the offense rebounded. Then it averages the team's count with its opponent's to steady it.
For a player, the name covers two different numbers. Basketball-Reference uses Dean Oliver's individual rating, which its explainer sums up in Oliver's words: "How many points is a player likely to generate when he tries?" It credits a player for the points he produces, including a share of the baskets he assists, and it charges him a possession for every turnover and every miss his team doesn't rebound. NBA.com's player offensive rating is his team's points per 100 possessions while he's on the floor, no matter who scores them.
Oliver also warned that his rating has to be read next to usage. The bigger a player's role, the harder it is to keep the number high, so a spot-up shooter who only takes open looks will often out-rate the star who creates them. The shooting half of that story, and why field goal percentage gets it wrong, is in the post on true shooting.
Per-100 numbers take pace out of the picture
Teams don't play at the same speed. In 2025-26, Miami averaged 103.4 possessions per 48 minutes, the fastest pace in the league, and Boston averaged 94.8, the slowest, per Basketball-Reference's 2025-26 league page. That's almost nine extra trips a night for Miami, and it warps anything you count per game.
Miami scored 120.9 points per game, second in the NBA, while Boston's 114.9 ranked 19th. Per 100 possessions, it flips. Boston's offensive rating of 120.8 was second in the league, and Miami's 116.7 was 13th. The slow pace flattered Boston's defense the same way: it allowed the fewest points per game in the league (107.2) and ranked fourth by defensive rating (112.7).
What plus-minus means in basketball
Basketball-Reference defines plus-minus as "the difference between the team's scoring and the opponent's scoring while the player is on the court." It doesn't care who scored. If your team goes on a 10-2 run while you stand in the corner, you're +8 for that stretch.
To calculate it, add up his stints. Say a player checks in with his team down 20-24 and checks out with it up 35-31. His team scored 15 and allowed 7 while he played, so he's +8. He checks back in at 60-58 and stays on until the final buzzer of a 101-99 win. That stretch went 41-41, so he's even for it and +8 for the game.
There are always five players on the floor, so a team's plus-minus numbers always sum to five times its final margin. Take the night Devin Booker scored 70 at Boston in March 2017. The Suns lost 130-120, and the box score shows their eight players' plus-minus adding up to −50. Booker finished −6 on his 70 points, while Jared Dudley, who scored six, finished +14. Basketball-Reference even gave Booker's own possessions a 131 offensive rating that night, against 109.8 for the Suns as a team. He was brilliant, and his team still lost his minutes by six.
A player's plus-minus includes his teammates
Raw plus-minus gives all five players on the floor the same credit, so a player's number drifts toward his teammates'. The per-100 version, on-court net rating (NBA.com lists it as a player's net rating), has the same problem. A better check is on-off: compare the team with him on the floor to the team with him on the bench.
In Basketball-Reference's 2025-26 play-by-play numbers, these three players had nearly identical on-court marks and very different on-off numbers:
| Player (team) | On court, per 100 | On minus off |
|---|---|---|
| Isaiah Joe (Oklahoma City) | +10.8 | −0.7 |
| Nikola Jokić (Denver) | +10.7 | +13.5 |
| Moussa Diabaté (Charlotte) | +10.4 | +10.7 |
Oklahoma City was +11.5 per 100 with Joe on the bench, a hair better than with him on the floor, so his +10.8 mostly measured the Thunder. Denver got outscored by 2.8 per 100 with Jokić on the bench.
Box plus-minus takes a different route. Basketball-Reference's BPM works from the box score to estimate how many points per 100 possessions a player adds over a league-average player. For a single player's night, though, minutes and role still matter more than any plus-minus, and that's the ground how to predict player stats covers.
Net rating and the record can disagree
A win by one point and a win by 20 count the same in the standings. Net rating counts the 20.
The Los Angeles Lakers went 53-29 in 2025-26 with a net rating of +1.8, 14th in the league. The Charlotte Hornets went 44-38 at +5.0, eighth in the league and fourth in the East, and missed the playoffs. Basketball-Reference's expected records, built from each team's scoring margin, nearly swapped them: Charlotte at 53-29 and the Lakers at 45-37.
Close games made the difference. The Lakers went 25-8 in games decided by fewer than 10 points, and Charlotte went 12-19, per the expanded standings. In games decided by 10 or more, Charlotte had the better record, 32-19 to 28-21.
Some teams do keep winning close ones. The Lakers also beat their expected record the season before, going 50-32 on a +1.2 net rating. Still, if you want to know how well a team has played, net rating is the cleaner read, because a handful of late possessions can't swing it much.
How net rating would change a pick
Say a basketball-style pick in GAGE asks whether a team wins, and the odds give it 40% when you lock YES. Its record is worse than its opponent's, but its net rating is better, the Charlotte-and-Lakers kind of split. That pick would be worth +6.0 if the team wins and cost 4.0 if it loses. If you're not sure what a 40% chance is telling you, start with how to read a favorite and an underdog.
The full table is on how GAGE scores a pick, and GAGE is on the App Store when you want your own reads scored.