(Image via Bleacher Report)
Player Efficiency Rating (PER) was the turning point for basketball analytics in the public domain; but with time comes controversy. Lots will claim that PER is an outdated metric, that recent efforts of all-in-one metrics have left PER useless. Let’s say this is true, and for good reason. PER has a limited scope compared to lots of modern metrics, which incorporated Plus/Minus and tracking data that seems to map a player’s value to an objectively greater degree than the box score. Let us be reminded that this is a good thing! More advancements means more descriptive power among these all-in-one metrics, so the science of player analysis has undergone significant evolution since Hollinger’s initial effort. So, as a brief summation, PER is NOT one of the absolute best measures of player impact for the modern era. But like any other subject with a rich history, there is value in going back to appreciate tradition. Let us explore the inner workings of PER. What makes it good or bad, and why should or shouldn’t you use it?
How PER Works
PER puts the “E” in “Efficiency” because it was designed as a fairer comparison between high-volume superstars and role players. Like lots of modern impact metrics, PER is a per-possession metric that estimates a player’s effect on his team’s point differential . But instead of using advanced, complex mathematical modeling to accomplish its results, PER is grounded in concept and critical thinking. It works so that each box score statistic (points, assists, etc.) is assigned a weighting, which is personalized to its box score stat such that each weighting reflects the statistic’s point-value. If that sounds a bit confusing, hopefully this example makes things clearer:
LeBron James committed 196 turnovers  during the 2022 regular season. Most basketball fans will agree that turnovers are bad! A turnover means the offensive team lost possession of the ball without having scored. Thus, they lost out on the opportunity to score points in that possession; and basketball is often a game that is won in a matter of possessions. Plus, there’s also the drawback from permitting the opposition a transition possession, which tends to yield more points than typical half-court offense. So we can all agree turnovers are bad, right? But how do we know how bad one is? Well, we have to consider what a turnover is taking away from, and the big thing is those aforementioned points the team could have scored if the ball weren’t turned over.
Since a variety of outcomes could’ve followed barring one of James’s turnovers, Hollinger decided to use a single placeholder number to sum up all possible outcomes of a possession. This estimates just how many points the Lakers were missing out on because of the turnover. Let’s think about it this way: on average, NBA offenses scored 1.12 points per possession in the 2022 regular season . Hollinger concludes this was the best estimate of how much an offense is missing out on from turning the ball over, not accounting for how the opponent can gain a transition possession, as stated earlier. But since the turnover is deducting that amount of value, a turnover in 2022 is worth approximately -1.12 points! If we apply that weighting to all of LeBron James’s 196 turnovers this year, we can estimate that he cost the Lakers almost 220 points from his turnovers.
This process is done for all box score statistics (most of which have far more complicated weightings), and the results are summed up to estimate how many points a player is adding (or subtracting) to his team. However, as stated earlier, there is one last adjustment that converts the results to a per-possession basis. Namely, superstars don’t receive any benefits for playing more of the game. This “adjusted” form of PER is what you’ll see on Basketball-Reference or ESPN leaderboards. With that out of the way, let’s discuss the good and the bad of PER, and what its process means for the metric’s ability to evaluate players.
The Good and the Bad
Believe it or not, there is some good stuff about PER. Take this as more of an opinion than a fact, but PER has solid (and interesting) critique behind it. The idea that each statistics has a hypothetical point-value behind it is logic that is used to this day, even with the most revered metrics like Estimated Plus/Minus and BBall-Index’s LEBRON; and PER was the first to do it. Let’s also consider the benefit to translating the results to per-possession rather than per-game or across an entire season. PER can be a reasonable tool for identifying hidden gems slotted at the end of the bench, and can act as an indicator of whether a role player is outperforming one of the better players on the team. Back when PER was the gold standard, there was some value it could provide to teams looking to manage their roster and shuffle their lineups. Not at all a bad result!
However, there are a considerable number of drawbacks to discuss with PER; and unfortunately, these dominate the conversation and are often valid concerns of the metric. Let’s return to the per-possession basis of PER. While there may be benefits, there are also flaws with this approach. There is the matter of fatigue in which players with a heavier load tend to tire out near the end of games, whereas role players can spend more of their time on the court with a fresh set of legs. Superstars also tend to player with better teammates, meaning they have fewer chances to exert their skill; and they often play tougher opponents, whereas role players tend to match up with the opponent’s bench units as well. That is, not all statistics are created equally, and a player’s impact can be reframed and reimaged depending on who else is on the court, which is something PER struggles to account for.
The other glaring issue with PER is that it lacks a benchmark, or some type of measurement to fall back on to confirm the metric is grounded in reliability. If you’re familiar with other popular impact metrics, you may remember that they are often backed by RAPM, which is arguably the strongest all-in-one metric basketball has to offer. If not, I have written in-depth about RAPM before  if you’re interested in learning more about it. (TW: Math.) PER has been around for nearly twenty years, so it’s no surprise that it has been one-upped by more recent metrics, but the larger issue is that John Hollinger has refused to update the metric. PER has been more or less the same since 2005, which excludes the metric from the comprehensive analytical techniques of today. This, paired with the inconclusive nature of the metric’s logic, makes PER one of the less viable options to analyze players.
Should You Use It?
The biggest argument against using PER is that there are many better alternatives, which is true. Estimated Plus/Minus, LEBRON, RAPTOR, and BPM, among others, all have more solid bases in accomplishing their goals. We’ve discussed the positives and the negatives of the metric, and are hopefully all able to recognize that both exist. But, again, PER is simply outdated. Hollinger’s refusal to refine his metric has left it in the dust. You’re much better off using the Backpicks (proprietary) or Basketball-Reference models of BPM if you’re looking to view a player’s impact from the perspective of the box score. But does this mean PER is bad?
I wouldn’t say so. PER has a very intriguing basis, and the logic Hollinger incorporated into the metric shows a considerable understanding of how basketball is played and won. Therefore, PER is still going to be quite a hit-or-miss metric, as lots of them are. But it’s not “bad” to the point that it measures every player off the mark. Heck, there’s the off-chance that PER pins a player down just right! However, it’s not great either, especially considering its peers. So the next time someone tells you PER is the worst thing to grace the Earth, tell them it could be worse. If it helps, tell them to look into Wins Produced from the Wages of Wins Journal. But that’s all from me, folks. I hope this was a somewhat insightful look into PER, and that you’re next perusing of a leaderboard is that much more confident.
Leave a Reply