Breaking Down the NBA’s PER Metric

Welcome to the second installment of our basic analytics primer here at Basketball Insiders, where our goal is to give a general education on many of the simplest and most common “advanced” metrics used by statisticians in the modern NBA. Last week, we covered the various forms of plus-minus metrics, from raw on-court numbers to much more advanced figures with high-level mathematics involved.

This week, we move on to one of the first and most initially progressive metrics on the scene: PER. Let’s get right to it.

What is PER?

PER is an acronym, for those a bit slow on the uptake, and it stands for Player Efficiency Rating. Developed by then-ESPN statistician John Hollinger, who now holds a central role in the Memphis Grizzlies’ front office in part due to his considerable success as a curator and creator of statistics, PER is a single number that attempts to combine all the relevant statistical contributions a player makes. The primary goal isn’t to determine raw volume, but rather to quantify how efficient a player is in every countable aspect of the game.

The formula for PER is quite complex, and includes everything from field goals made and missed to steals and turnovers. The eventual output is a per-minute measure and is pace-adjusted – meaning guys who play for teams that slog it up won’t be penalized for having fewer possessions with which to put up numbers. Perhaps the part of PER that made it so immediately popular and widely used was the fact that it has a league average figure that remains the same every year: a 15.0 PER is the mean, the baseline against which all other players stack up. A guy below 10 is therefore a significantly sub par player by this metric, while anyone over 20 is doing very well and approaching star status in short order.

PER is a very easy statistic to find for anyone with an internet connection. It’s housed at several major sites, including its original home at ESPN. The ease of access and the simplicity of the single-number output have made PER perhaps the most commonly used “tell-all” metric in the game.

What Can We Learn From PER?

PER is an excellent way to measure overall efficiency, particularly on the offensive end where the base statistics offer a large number of data points. The outputs tend to conform very well with what our general perceptions of efficiency are. When LeBron James was dominating the league both in his first Cleveland Cavaliers stint and then with the Miami HEAT, where he topped 55 percent shooting with insane box score stats on a couple occasions, his PER was among the highest in league history (his 27.66 career mark is second in NBA history, behind only Michael Jordan).

It’s not a one-size-fits-all metric by any means (more on this in a little bit), but PER has traditionally done a pretty darn good job at recognizing whom the popular consensus would consider the league’s best players. Last year’s top eight for PER, in order, were Anthony Davis, Russell Westbrook, Stephen Curry, James Harden, Chris Paul, LeBron James and Blake Griffin. One can quibble with the order they’re in and perhaps do the same as they move further down the list, but there wasn’t a single guy in last year’s top 20 who wasn’t, at minimum, one of the top offensive players in the league at their position.

Now, one of the chief drawbacks of PER is that it doesn’t capture much on the defensive end. As a box score-reliant measure, there are just very few data points it can touch that will give an accurate representation of how well a guy plays defensively. Things like steals and blocks are only partially useful. For this reason, PER is not advisable as a metric for evaluating guys on that end of the floor, and there remain few such metrics (though certain advanced plus-minus models do a pretty good job, as we discussed last week).

Who Are Some PER Outliers?

Both because of the defensive stuff noted above and due to small sample sizes or curious usage in certain cases, PER will have outliers from time to time on either end of the spectrum. Jeremy Evans, formerly of the Utah Jazz and now with the Dallas Mavericks, has always been one of these – he’s never been among the elite for this figure but, for instance, finished 38^th last year for PER. It’s not the first time he’s sported a well above-average figure despite playing very rarely over a full season, and it’s likely that he’s one of the guys who has been bad enough defensively that the metric misses a big negative area.

Evans is one of very few true outliers among guys with any reasonable sample of minutes, though. Sim Bhullar actually led the NBA in PER last year, but everyone understood it was a spurious title because he only played three total minutes on the year. One example is Brandan Wright, who had a 20.44 PER last year and finished the season 34th in the NBA (above All-Stars like John Wall and Paul Millsap among others) while playing for three different teams during the 2014-15 campaign. This was largely due to his very high field goal percentage. Guys like Evans and Wright, who have actually played a three-digit (or four-digit) number of minutes but still wildly differ in PER from what their consensus quality is, are much rarer.

Another Utah player highlights the other end of the spectrum: Dante Exum was a big minus offensively in his rookie season for the Jazz, which PER picked up while assigning him a 5.7 figure – worst in the NBA among guys who played over 1,000 minutes last year. But the Jazz had a ton of success with Exum on the court as their starter later in the year, and his on-court plus-minus figures were, while not elite, much better than his PER would suggest. Again, this mostly appears to be PER missing some elements of context, mostly on one end of the floor – Exum quickly developed into a dangerous defender at the point, and a guy who meshed well with the rest of Utah’s starting group on both ends of the floor. He didn’t deserve to be rated as a star or even at average level as a 19-year-old, but there’s also an easy argument that he wasn’t among the worst high-volume players in the league, as PER would suggest.

As with any single big-picture metric like this, use caution before leaning on it too heavily. The drawbacks listed above are legitimate, and even when only looking at areas PER tracks well, there can be holes. PER and all other box score metrics offer absolutely nothing as far as teammate chemistry or team scheme, which is often a huge part of what makes a guy valuable to a given group.

When used correctly, though, PER is quite helpful. It can summarize a guy’s efficiency in a simple way that’s easy to compare both presently and historically, and does a great job mirroring what the consensus opinion would be in this area. It will remain one of the baseline staples of single metrics.