Player Efficiency Rating (PER) Explainer
Player Efficiency Rating (PER) Explainer
What John Hollinger’s player efficiency rating measures in basketball, what it needs that a box score does not contain, and what it is criticised for. This page deliberately computes no PER, and the reason is data rather than caution.
What PER needs, and what you can use instead
Explainer, not a calculatorTrying to work out a player’s PER from the player’s own box score line
What the formula needs, and where each input lives
| Input | On a player’s box score line? | What it is for |
|---|---|---|
| The player’s points, field goals, attempts, threes, free throws, rebounds, assists, steals, blocks, fouls, turnovers and minutes | Yes | The events being scored |
| Team assists and team field goals | No, team season totals | How much credit a basket earns when somebody else created it |
| Team pace | No, a derived team figure | The pace adjustment: a fast team’s players must not be flattered |
| League assists, field goals and free throws | No, league season totals | Forming the factor that sets the assist and basket weights |
| League points, field goal attempts, offensive rebounds, turnovers and free throw attempts | No, league season totals | Forming the value of a possession |
| League total and offensive rebounds, and league pace | No, league season totals | The league defensive rebound share, and the pace adjustment |
| The league’s minutes-weighted average adjusted rating | No, and not derivable from any one line, team or raw league total | The final rescaling that puts the league mean at fifteen |
Four composite ratings, checked the same way
| Rating | Is the method published? | Could it be reproduced freely? | Computable from a box score? |
|---|---|---|---|
| Player efficiency rating | Yes, in full as plain text on Basketball-Reference, which credits John Hollinger and his books | Could not establish that it is free. No licence, fee schedule or permission statement was found, and no copyright notice specific to it | No. Needs league season totals, league and team pace, and the league’s own minutes-weighted average rating |
| Win Shares | Largely, on Basketball-Reference, which defers the post-1977 offensive method and the defensive rating to Oliver’s book | Could not establish that it is free. No licence or reuse statement for the method was found | No. Needs league points per possession, league pace, league points per shot attempt, team pace and team defensive possessions |
| Box Plus/Minus, version 2.0 | Yes, by its author Daniel Myers, with the regression coefficients published as tables | Could not establish that it is free. No licence or reuse statement for the method was found | No. Needs team adjusted efficiency per 100 possessions, league regression baselines, and estimated position and offensive role |
| Value Over Replacement Player | Yes, as “[BPM – (-2.0)] * (% of possessions played) * (team games/82)” | Could not establish that it is free, and it inherits Box Plus/Minus | No. It is a transformation of Box Plus/Minus and inherits everything that one needs |
What PER is for, why this page will not compute one, and what it is criticised for
This page is about basketball, the five-a-side game played to a basket ten feet above the floor. It has no single rule-maker: the Fédération Internationale de Basketball (FIBA) publishes the Official Basketball Rules used by most of the world, the NBA publishes its own rulebook for its own competition, and the NCAA publishes separate rules for United States college play. Where a number on this page is one of theirs rather than plain arithmetic, the body and the edition are named beside it.
The player efficiency rating is John Hollinger’s attempt at one number for a player’s statistical production, and the appeal is obvious: a box score has a dozen columns, a reader wants a ranking, and PER provides one on a scale that is easy to carry around, because the league average is fifteen by construction every season. In outline the method credits every positive event in a player’s line and debits every negative one, with weights built from league-average scoring values rather than chosen by hand; puts the total on a per-minute basis so a reserve and a starter are comparable; adjusts by the ratio of league pace to the player’s team pace, so a fast team’s players are not flattered by playing in more possessions; and finally rescales the whole league so the average lands on fifteen.
The obstacle is not arithmetic but data. The formula needs league-wide season totals — the league’s assists, field goals, free throws, points, attempts, offensive rebounds, turnovers and free throw attempts, the league’s pace, the team’s pace, and finally the league’s own minutes-weighted average rating, because the last step divides by it to set the league mean to exactly 15.0. None of that is on the box score in front of you, and the last cannot be worked out from any one player’s line at all: it depends on every other player in the league. The first few missing pieces are inconvenient — a league’s season totals are published somewhere, even if not where the reader is standing. The last is different in kind. The league’s minutes-weighted average rating is a function of every player in the league: to get it you must compute the adjusted rating for all of them and weight by minutes. There is no table to look up and no constant to hard-code, and a fitted stand-in would fail exactly where it matters, because a reader checking a rating is almost always checking a player near a threshold rather than one in the middle.
There is a second finding, deliberately recorded as the weaker of the two kinds this site distinguishes. It could not be established that the formula is free to reproduce. Basketball-Reference prints the whole expression as plain text and attributes it to Hollinger, noting that he “writes and sells books” and that the method was learned from them; no licence, fee schedule, permission statement or copyright notice specific to the rating was found anywhere, and Wikipedia’s article on it says nothing about its status. That is not the same as establishing that it is encumbered, which would need a licence or a fee schedule actually read, and this page does not make that claim. The data wall above stands on its own and would be reason enough by itself. The same two checks were run on Win Shares, Box Plus/Minus and Value Over Replacement Player, with the same result in both columns, and none of them is built here either.
The criticisms are worth printing rather than summarising away, because a reader handed a PER needs them. The sharpest comes from Hollinger himself, who has said the rating “is not a reliable measure of a player’s defensive acumen” and that blocks and steals — the only defensive events in a box score — can distort a player’s value, which is why a defensive specialist can post a single-digit rating through a long and valuable career. David Berri argued in The Wages of Wins that “a player can be an inefficient scorer and simply inflate his value by taking a large number of shots”; Hollinger answered that this was “a huge misunderstanding of PER” on the grounds that the break-even points are set by league calibration. It is also commonly said that the rating flatters players in limited minutes and against second units, which is the same problem the per game and per 36 minutes calculator is written about. And the normalisation that makes the scale readable also seals it: a rating is a statement about one league in one season and does not travel.
What to use instead, if what you wanted was a defensible efficiency figure from the numbers you have. The true shooting percentage calculator prices all three ways of scoring from three box score fields; the effective field goal percentage calculator does the three-point weighting with no estimated coefficient in it at all; and the offensive and defensive rating calculator gives points per 100 possessions from a count you supply, with the possessions and pace calculator behind it. None is one number for a player, and that is the honest position: the single number is where the league-wide constants and the chosen weights come in.
Frequently asked questions
Can I calculate PER myself from a box score?
No. Before a single event in a player’s line can be scored the formula needs three factors built from the league’s season totals, then the league’s and the team’s pace, and finally the league’s own minutes-weighted average rating — which is a function of every other player in the league. A box score line contains none of that.
Why does this page not have a PER calculator?
Two independent reasons. The formula needs league-wide seasonal constants that are not on the page in front of you, and the last of them cannot be derived from any one line at all. And it could not be established that the formula is free to reproduce — a weaker finding than saying it is encumbered, and recorded as such. Either reason alone would be enough.
Why is the league average always 15?
Because the final step divides by the league’s own average rating and multiplies by fifteen, so the mean lands on fifteen every season by construction. That is what makes the scale easy to read and what makes two seasons incomparable: a rating says “this far above this league’s average this season”, and nothing about another league or another year.
What is PER criticised for?
Chiefly for defence. Hollinger himself has said it “is not a reliable measure of a player’s defensive acumen”, because the only defensive events in a box score are blocks and steals. David Berri argued in The Wages of Wins that an inefficient scorer can inflate the rating by taking many shots, which Hollinger called “a huge misunderstanding of PER”. It is also said to flatter players in limited minutes.
What about Win Shares, BPM and VORP?
All three were checked the same way and none is built here. Each method is published — Box Plus/Minus with its regression coefficients in full — and for each it could not be established that reproducing it is permitted. More decisively, each needs team and league-wide inputs a reader does not have: team adjusted efficiency per 100 possessions, league baselines, league pace, estimated position and role.
What should I use instead of PER?
True shooting percentage for scoring efficiency, effective field goal percentage if you want a figure with no estimated coefficient in it at all, and offensive or defensive rating per 100 possessions for a team. All three are published definitions computable from numbers you already have. None is a single all-in-one number, which is the trade rather than a shortcoming.
Related calculators
References
- Sports Reference LLC. “Calculating PER.”, read 12 Oct 2026. Describes PER as “a per-minute rating developed by ESPN.com columnist John Hollinger”, learned from “his books”, noting Hollinger “writes and sells books”. Publishes the unadjusted formula in full as plain text with its three league constants, the pace adjustment lg_Pace / team_Pace, and the normalisation PER = aPER * (15 / lg_aPER), lg_aPER being the league’s minutes-weighted mean. Carries no licence, permission statement or copyright notice specific to PER.
- Wikipedia. Player efficiency rating, read 12 Oct 2026. The weakest source here, cited for three things: that its sentence “The formula is:” is followed in the fetched text by nothing at all, the equation existing there only as a rendered image; that it carries no statement about PER being proprietary, copyrighted, trademarked or licensed; and for Hollinger’s own stated limits and David Berri’s criticism, both quoted on this page.
- Sports Reference LLC. “Calculating Win Shares.” Written in the first person; the author describes “stealing a page from baseball’s Bill James” and defers the post-1977 offensive method and the defensive rating to Dean Oliver’s Basketball on Paper. Requires league points per possession, league pace, league points per shot attempt, team pace, team and opponent points per game, team defensive possessions and team minutes — none of it on a player’s box score line. No licence or reuse statement for the method. Read 12 Oct 2026.
- Myers D. “Box Plus/Minus, Version 2.0.” basketball-reference.com, February 2020, read 12 Oct 2026; bylined “by Daniel Myers, developer of Box Plus/Minus”. Publishes the regression coefficients as tables. Beyond a player’s box score line it needs team adjusted efficiency per 100 possessions, league regression baselines, team-mates’ minutes-weighted raw BPMs, and estimated position and offensive role. VORP is “[BPM – (-2.0)] * (% of possessions played) * (team games/82)”. No licence or reuse statement for the method.
- Sports Reference LLC. “Calculating Individual Offensive and Defensive Ratings.” States that the ratings were “developed by Dean Oliver in his 2004 book Basketball on Paper” and refers readers to the book for full details. The individual ratings need the team’s full season line, the opponent’s made and attempted field goals and free throws, points, turnovers and offensive rebounds, and team possessions. No team possession formula appears on the page. Read 12 Oct 2026.
- Sports Reference LLC. Basketball-Reference.com Glossary, read 12 Oct 2026. True shooting attempts “FGA + 0.44 * FTA”, eFG% “(FG + 0.5 * 3P) / FGA”, TOV% “100 * TOV / (FGA + 0.44 * FTA + TOV)”, Pace “48 * ((Tm Poss + Opp Poss) / (2 * (Tm MP / 5)))”. Its team Possessions entry instead uses 0.4 on free throw attempts and 1.07 on an estimated offensive-rebound term, and averages both teams’ estimates — so one page carries 0.4 and 0.44 for the same quantity. Offensive and defensive rating carry no formula here.