PER, Explained: What Hollinger's All-in-One Metric Gets Right and Wrong

For the better part of two decades, if you wanted one number to summarize a basketball player, you reached for PER. John Hollinger's Player Efficiency Rating was the first all-in-one stat to go mainstream — the number that let a fan say "he's a 25-PER player" and have everyone in the room nod. It's still everywhere, still quoted, still on the front of every player page. And it's still, in important ways, wrong about the things basketball cares about most. Understanding exactly what PER nails and exactly where it goes off the rails is the best possible education in why the metrics that came after it had to exist.

What PER is actually built to do

PER sets out to take everything a player does in the box score — every positive act and every negative one — and crush it into a single per-minute number. Three design choices define it, and all three are genuinely smart.

First, it's per-minute, not per-game. A bench player who produces in 18 minutes isn't punished for sitting; he's measured at the same rate as a starter. Second, it's pace-adjusted, so a player on a run-and-gun team that takes 105 possessions a night isn't artificially inflated over one grinding out 92 — the extra raw counting stats that come from extra possessions get normalized away. Third, and most famously, it's league-normalized to a fixed scale every single season: the league average PER is set to exactly 15.00, every year, by construction.

That last choice is the source of PER's staying power. Because 15.00 is always average, the scale is portable across eras. A 20 means the same thing — comfortably above average — whether it was posted in 2004 or 2026. The rough reference points everyone internalizes flow directly from that anchor.

The PER scale (league average is fixed at 15.00 every season)~15 = league average · ~18–20 = solid starter · ~22–25 = All-Star level · ~27.5+ = MVP candidate · ~9 = end of the bench

How it's constructed, conceptually

You do not need the full formula to understand PER, and frankly the full formula obscures more than it reveals. Conceptually, it works in a few steps. Start by adding up everything good a player did that the box score records — points, made field goals and free throws, rebounds, assists, steals, blocks — each weighted by an estimate of how much that act is worth in points. Then subtract the bad — missed shots, turnovers, fouls, each weighted by its cost. That gives a raw value of positive contributions minus negatives.

Next, divide by minutes to get a per-minute rate, and apply the pace adjustment so teams with more possessions don't get a free boost. Finally, scale the whole league so the average lands on 15.00. That last rescaling is what turns a raw, hard-to-interpret per-minute sum into the familiar number on the player page.

PER, in one line of plain EnglishPER ≈ (sum of weighted positive box-score acts) − (sum of weighted negative acts), expressed per minute, pace-adjusted, then rescaled so league average = 15.00

The weights themselves come from Hollinger's accounting of how many points each event is worth on average — the same lineage of possession-value reasoning that underpins Dean Oliver's work. The machinery is more involved than this, but every bit of the extra complexity is in service of those weights and adjustments, not a different idea.

What PER gets right

Used for what it's good at, PER is excellent. It is a fast, single-number summary that is genuinely minutes-proof and pace-proof — two things raw per-game stats badly fail at. It's available for essentially every player in NBA history, because it needs only the box score, which means you can compare across eras on a stable 15.00-centered scale. As a first-pass sorting tool — drop two hundred players into a sane order in one column — it does the job, and for offensive-minded, high-production players it tracks intuition well. If all you want is a quick read on a high-usage scorer's overall box-score output, PER will rarely embarrass you.

Where PER goes wrong — and it goes wrong in knowable ways

The biases in PER are not mysterious; they're structural, and Hollinger himself acknowledged most of them. They all flow from the fact that PER is built almost entirely from positive box-score events, which means it rewards doing things — and doing things requires having the ball.

The first and most cited flaw: PER rewards high-volume scoring in a way that doesn't fully account for inefficiency. Because made shots add value and the penalty for a miss is comparatively gentle, a player can pad his PER by simply taking lots of shots, even at mediocre efficiency. The metric credits the makes generously and forgives the misses too lightly, so a chucker can post a respectable PER that his actual scoring efficiency doesn't deserve. This is exactly the failure that pairing the stat with True Shooting percentage and usage rate exposes — PER can't, on its own, tell you whether the volume was worth it.

The second flaw is the flip side: PER undervalues the value of not shooting. The smart pass-up — declining a contested shot so a teammate can take a better one — is one of the most valuable things an offensive player can do, and PER literally cannot see it, because nothing was recorded. A player whose discipline lifts everyone around him looks, to PER, like he simply did less.

The third, and the one that matters most: PER badly undervalues defense. Because the only defensive events in the box score are steals, blocks, and defensive rebounds, PER's entire picture of defense is those three columns — and as covered in the all-in-one metrics overview, those columns reward gambling and miss positioning entirely. A great positional defender who deters shots and never gambles gets almost no defensive credit in PER. An elite rim-protecting, never-leave-your-feet anchor can look pedestrian; a ball-hawk who gambles for steals can look better than he defends.

A clearly-illustrative example

Imagine two forwards who play identical minutes. Forward A takes a heavy volume of shots at middling efficiency, grabs a healthy pile of rebounds, and gambles for a couple of steals a night — but his man scores freely and his contested-shot defense is poor. Forward B takes fewer shots but only good ones, passes up contested looks to create better ones for teammates, records almost no steals because he never leaves his assignment, and quietly locks down the opponent's best scorer every night.

PER will very likely rank Forward A meaningfully ahead of Forward B. The volume scoring, the rebounds, and the steals all feed PER directly; the disciplined shot selection and the lockdown defense feed it almost nothing. Anyone who has watched both players knows the ranking is backwards. That inversion — high-volume, stat-generating play floating to the top while efficient, defense-first play sinks — is PER's characteristic failure mode in a single picture. (Both forwards here are invented to illustrate the bias, not real players.)

Why the next generation of metrics exists

Every major all-in-one metric built after PER — Box Plus/Minus, EPM, the various plus-minus and hybrid approaches — was designed, in part, to fix one or more of PER's blind spots. Plus-minus methods don't care whether an act was recorded; they ask only what happened to the scoreboard while a player was on the floor, which finally lets the value of deterrence, smart pass-ups, and positional defense show up. Hybrids bolt that on/off signal onto box-score inputs to stabilize it. The lineage is direct: PER defined the genre and proved the appetite for one number, and its specific, well-understood failures became the design spec for everything that followed.

So keep PER for what it is: a fast, historically available, pace- and minutes-proof first read on box-score production, anchored to a scale that has meant the same thing for twenty years. Just never let it be the last word — especially on defense, on efficiency at volume, or on the players whose best contribution is the shot they chose not to take. PER tells you how much a player did. It was never built to tell you how much that doing was worth.