In this video, we’ll be discussing team offensive, defensive, and net rating in basketball. We’ll be offering our explanations of what the ratings are and how they’re typically used, while also briefly discussing some of the assumptions made when using the metrics.
As a formula, the ratings are not too complicated: Take the number of points scored for offensive rating or points allowed for defensive rating, divide by the estimated number of possessions played, in the given game or season, and multiply by 100. To create net rating, subtract defensive rating from offensive rating. This can be thought of as the point differential per 100 possessions. The 100 possession benchmark is used as a consistent and aesthetically pleasing baseline for the total number of possessions in an NBA game. With every team’s production placed on the same number of possessions per game, the metric is meant to even out the playing field when comparing different offenses and defenses.
So, what can these metrics be used for? Well, there are generally three levels of analysis for team sports like basketball: team-level, rotation-level, and player-level. Team level analysis lets you compare team performances, rotation-level analysis lets you compare the performance of certain groups and matchups of players, and player-level analysis lets you compare individual performances. These metrics are most commonly used at the team and rotation level. There are separate calculations for an individual player’s offensive and defensive rating. Typically, the ratings are used for general comparison, but when analyzed properly can also offer comparisons across eras. It is no secret that the pace of play in the NBA has radically increased over the last several years, and what these metrics do, by normalizing to 100 possessions, makes it so performance by these measures can be judged regardless of pace.
Pace in this context refers to the number of possessions in a game. The faster you play, the more possessions there will be by BOTH sides. The slower you play, the fewer possessions both sides will have. For example, the idea is that, with net rating, we could compare the performance of the grit-and-grind grizzlies and the pace-and-space warriors with ease.
Of course, that ease is courtesy of a set of assumptions about how to use the ratings. As with any other sports metrics, you don’t have to use offensive, defensive, or net rating if you aren’t comfortable with the assumptions being made, but let’s dive into the three major assumptions that go into the metrics.
The first fundamental assumption being made is that possessions can be estimated accurately enough through box score stats. The pro for this assumption is that it estimates it the same way for everybody, so it wouldn’t be doing anyone in particular any favors. Also, it is a reasonably close estimation. The con is that it tends to overestimate possession totals. With play-by-play data, we can actually count the number of possessions in a game, and Justin Jacobs, in a post on his squared2020 blog a few years ago, compared the number of possessions this method estimates to have happened to the actual number. The result? That it consistently overestimates. So, it’s the same for everybody, but also a little bit high for everybody.
The second major assumption being made is that points are the best judge of a team’s performance. This may sound silly, but it’s worth thinking about. Points decide who wins and loses, so it is perfectly acceptable means of measurement. But point totals can be noisy, meaning that there can be oddities that go into how a point total in a given game is reached. Every shot behind the three point line is worth three points, whether a full court heave at the buzzer, a catch and shoot three pointer from a drawn up play out of a timeout, or anything in between. Even that Trevor Booker scoop shot as the shot clock expired from a few years ago (link in the description). If the goal of offensive rating is to measure the quality of the offense, it is worth considering this assumption based on the sample size of possessions contributing to the rating.
The third major assumption is that pace should NOT be considered part of team performance when measuring using this metric. This metric deliberately gets rid of pace from the naked eye. But what’s important to consider is that the fewer possessions a team has, the fewer opportunities there are to succeed or fail, which puts more pressure and mathematical weight on each possession. The opposite is true for teams that play at a faster pace, there is less pressure on each possession. This is a fundamental part of a team’s offensive and defensive decision making. So, the argument could be made that these ratings don’t just normalize the number of possessions, it normalizes the team’s strategies and decision making. The obvious pro of this assumption is similar to the last one: points decide who wins and loses, and to make general comparisons, it’s kind of necessary to get rid of what goes into the point totals from the naked eye. The con is that it could lead to you inappropriately comparing two teams because you’re not considering any of the key things that have made their ratings what they are. Whether that’s what you’re trying to do or what you want to avoid doing is up to you.
So, in summary, offensive, defensive, and net ratings show the number of points scored, allowed, and difference between the two per 100 possessions the team (or lineup) plays. This is a pace-neutralized metric that is helpful when comparing teams across eras and play-styles. But, it also comes with its own bag of assumptions, that you should definitely think about before really looking at it and/or trying to interpret results born from analyses that use them.