One of the core conceits of advanced stats is that they tend to be normalized to some kind of league average. Player Efficiency Rating (PER) is built so that league average, every year, is exactly 15. In theory, the average for Win Shares per 48 minutes should be .100, since a team that has 0.1 WS per position per game will get half a win—half a win plus half a loss equals a 41-41 record over 82 games.
Value Over Replacement Player is a different animal; a “replacement player” isn’t the same thing as a league-average player. It’s meant to represent a competent player called up from the G-League, and as such a team with zero VORP over an 82 game season would win something closer to 20 games, and 10 VORP for a team is the “league average” 41-win team.
Over the next three Thursdays, we’ll take a look at all three stats, pull some numbers from the 2018-19 and 2019-20 seasons (it’s still way too early to try and track league-average players for 2020-21; we might circle back to this during the offseason), and see if we can find names and statistical patterns that can identify just what’s meant by a “league-average” NBA player and whether that player’s a fringe starter, a sixth man, or actually far better (or worse) than that expectation.
We begin with PER, because that’s the easiest one to look up. Pull up Basketball Reference’s Stathead tools, search for all the guys who played at least 1000 minutes with a PER between 14.8 and 15.2 (trying to narrow this any further simply doesn’t return enough names), and let ‘er rip. Ready? Go!
The List
In the 2019 and 2020 seasons, here’s the complete list of 17 guys with a PER between 14.8 and 15.2, sorted by descending WS/48 (in parentheses):
Derrick Jones ’20 (.153)
Rondae Hollis-Jefferson ’20 (.146)
Brook Lopez ’20 (.145)
J.J. Redick ’19 (.118)
Monte Morris ’20 (.112)
Myles Turner ’20 (.111)
Reggie Jackson ’19 (.105)
Jalen Brunson ’20 (.101)
Derrick White ’19 (.101)
Jayson Tatum ’19 (.097)
Aaron Gordon ’19 (.093)
Kelly Oubre ’20 (.093)
Aaron Gordon ’20 (.087)
Bogdan Bogdanovic ’20 (.084)
Caris LeVert ’19 (.077)
Terry Rozier ’20 (.066)
Kelly Oubre ’19 (.054)
The Patterns (?)
For one thing, the median WS/48 on this list is .101, and even though there’s wide variance between the best and the worst numbers at the extremes, the overall arithmetic mean is .1025. These are close enough to suggest that there’s at least some correlation between a league-average PER and a league-average WS/48 player, and the convergence is somewhere between the 2019 versions of Aaron Gordon and Reggie Jackson, while the archetypal 2020 “league-average player” was somewhere between Gordon and Monte Morris.
Interestingly, even though WS/48 tends to skew toward big men, Gordon failed to crack .100 WS/48 in either of the last two years. Volume shooters tend to get hurt by advanced stats, and Oubre is the prime example of that particular problem; so’s Terry Rozier.
But then again, the leader in WS/48 among 15-PER guys is a small forward. So there’s no real predictive value by position here.
The Problem
Therein lies the problem. The data’s way too noisy. You can’t draw a bead on what exactly PER is trying to tell us constitutes an “average” NBA player. There are inefficient shooters who get their counting stats elsewhere (Oubre and Gordon.) There are big men who are poor rebounders (Turner, Lopez.) And there are guys like Hollis-Jefferson, who shot himself out of the league but rescued his PER by virtue of not turning the ball over or missing too many free throws (RHJ’s free throw rate was over .400 last year, and he made 73.4 percent of them.)
We’ve already established that PER simply doesn’t correspond well with winning; it’s regarded, rightly, as something of a garbage advanced stat.
But we have our baseline. Derrick White was the “league-average” player in 2019, sitting exactly at the median of WS/48 for the 17-player sample of guys within 0.2 points of a 15 PER.
And Jalen Brunson was that guy last year.
White was a 1.3 VORP guy in San Antonio. Brunson was a 0.2 VORP guy in Dallas on a team that made the playoffs. So that’s no good either.
Tune in next Tuesday. I suspect we’re going to find a better representation of what league average means once we’re into stats that actually directly correlate with wins.
But if PER’s your stat of choice, there’s your answer.
