By Kurt Bullard
Ever since its inception, John Hollinger’s PER statistic has been a widely-used advanced metric that attempts to quantify a player’s overall performance. While the statistic has some practical use, one of its flaws—which John Hollinger admits to—is its trouble measuring defensive performance. It largely uses blocks and steals to do so, which does not paint a complete picture of a player’s contribution on the defensive end. While PER does a great job capturing offensive acumen, it fails to have the same accuracy on the defensive end.
However, PER still remains a very much cited advanced statistic to summarize overall play, not just offensive play. Doing so overestimates the worth of purely offensive players, while underestimate the worth of those whose main contributions come on the defensive end. One question I had was which players are most helped and most hurt by the use of PER as a statistic to measure overall performance by regressing PER against BPM—a statistic that better incorporates defensive play—and then see where the biggest discrepancies lie.
Using data from the 2016 NBA regular season, I ran this regression to see the relationship between BPM and PER. I regressed BPM against PER and PER^2, since there is not a requirement that the relationship between the two statistics has to be linear.
While the linear term is not significant according to the Wald test, an ANOVA test concludes that including PER as a predictor creates a better fitting model.
To see who is most harmed and most helped the use of PER against BPM, you can look at the residuals of the regression to see whether PER overestimates or underestimates the total worth of a player. A negative residual would suggest PER overestimates a player’s worth, since the actual BPM is lower than what PER would predict. The opposite holds true for positive residuals.
The following is a list of the players with the most negative residuals, suggesting that their worth is inflated by PER.
Most Overrated by PER
According to PER, the Thunder have three of the top ten players in the NBA this season: Russ, KD, and Enes Kanter. One of those three is very much unlike the others, and it’s not Westbrook or Durant. Kanter is great offensively, but a sieve on the defensive end. His PER is 24, tenth best in the association, but has BPM of -1.7, which is slightly better than replacement level in the NBA. Okafor has also been much maligned for his defense, both at Duke and with the Sixers.
The following is a list of the players who have the highest residuals.
Most Underrated by PER
Danny Green, who has a 10.4 PER (which is below the NBA average of 15), is the most maligned by the use of PER. The former Tar Heel has a BPM of 2.7, which ranks in the top 30 in the NBA. Bogut’s contributions lie mostly on the defensive end for the Dubs, while Draymond defensive impact is not captured by PER, even though Green has had a great offensive season.
While PER has its uses, I think everyone can agree that Enes Kanter is not a top 10 player in the NBA. Defense doesn’t get the love from PER that it truly deserves.
Great read! I feel PER also undervalues players that contribute invaluably to what their team needs. A few that come to my mind include Patty Mills, Cory Joseph, Jeremy Lin, etc. A statistic that could capture a player’s contribution to a team’s strengths would be nice.