Efficiency and Assignment

Those sound like basketball words, don't they?  I'm looking for a way to describe how much of a NBA team's offense is run through one player, and the good words are already taken.  Usage percent; feels right.  possessions used; a little bland but descriptive.  These are words and ideas that are established and useful but don't completely satisfy my notion of what a basketball role is or what it means to have a big role or small role on offense.  The name for my imaginary and not-yet-useful statistic that does is Assignment.

There are a lot of efficiency stats out there.  You've got your choice.  Some of them (annoyingly, to me) incorporate defense into the picture and create a mega efficiency stat.  For offense, you can look at a player's contributions independently - his shooting (with TS% or EFG%) turnovers (best captured by TOV%), offensive rebounding (OR Rate), and foul line trips (with FT rate or FTA/FGA+TO).  There are stats that combine all of these in one way or another - points per possession, offensive win shares, offensive WP48, floor percent.  The one I like to look at is offensive rating.

Offensive rating for a player is presented like points per 100 possessions, but it's a little more complicated than that.  It incorporates all of the above (shooting, offensive boards, free throws and TOs) and adjusts for assists.  It handles assists by splitting the credit for assisted baskets between the passer and scorer.  It does a pretty good job of  estimating a player's efficiency in the context of team efficiency.

According to ORtg, these and these are the all-time most efficient seasons for a player.  Most of the names make sense.  Great offensive players having great years, sometimes on teams that are sneaky-good on offense (like the 05-06 Pistons, led by Chauncey Billups and his 127 Ortg).  Shooters get recognized for the three-ball:  Reggie Miller was Mr. Efficient four times.  Unstoppable forwards like Barkley posted great years by this metric too.  Some names don't make immediate sense, though - like who the heck is James Donaldson?

In  1986-87, 7 foot 2 James Donaldson posted a league-high Ortg of 132 for the Dallas Mavericks.  He led the team in minutes, but you couldn't call him an offensive star.  He was just efficient with the few chances he got. 

The James Donaldsons of the NBA are why Ortg and all efficiency stats needs a companion statistic to measure how much role and responsibility the player has on offense.  In his book explaining Ortg, Dean Oliver uses possessions used, which has the downside for us of not being readily available on the internet.  The one that is available and useful is usage or usage%, which does not consider assists or who gets credit on assisted baskets.  This makes it slightly incongruous to pair it with Ortg, which does.  

Usage% is just how much of the offense one guy takes while he's on the court.  Think shots, free throws and turnovers, and recall that 20% of a team's possessions is the average and natural amount for one player to use.  We look at usage to know whether the player is a Kobe Bryant or a Bruce Bowen on offense; their tendency to assert, or defer, to shoulder the load or quite literally pass it on to a teammate. 

Donaldson's fit on the Mavs was to take very few possessions (10.5% usage), so despite his efficiency he didn't translate it into a lot of production.  A lot of the most efficient players have been on teams where they didn't have to carry the entire mission.

A more memorable example is Steve Kerr.  In '96 he posted an amazing mark of 141, the highest i've ever seen.  He was able to do this by only taking 12% of his team's possessions while on the court.  Micheal created the opportunities, Steve cashed 'em in. Kerr would have topped that Otrg list four times in a row but didn't meet the requirements for total points produced.

Oliver proposes a rough rule of thumb for how much a player can improve his efficiency by taking fewer possessions or hurt it by taking more than his limit.  Since we're talking in generalities we should stick to the outliers, the very high and very low volume guys.  So for limited players forced to make shots and decisions, and for star players passing on bad shots, the thumb rule says 1 percent of possessions used hurts/helps you by about one point of offensive rating.  Even more roughly we can say 1% usage translates to 0.01 points per possession, for stats that are available.

That Kerr and Donaldson had high-efficiency, high-usage teammates that allowed them to play the specialist role is obvious.  Their teams would have been great offensively with or without them, which is not something you can say about the less efficient, higher volume 'star' teammates.  The key is having the right role for the player and identifying who is versatile offensively and who is only efficient with a limited commission.  For different kinds of players there are different relationships between efficiency and responsibility, but the current theory is fewer shots means better shot selection.

This way we can see a player's contribution towards efficiency in light of his production.  Pau Gasol couldn't have dropped a 125 rating on the league in 09 if he still had to take 25% of the team's offense, like he did in Memphis.  Taking fewer shots generally means editing the tough looks out of your game. When a star volunteers to take 'the rest' of the shots it does help keep other guys in their sweet-spots, efficiency wise.  Your team usage leader is probably not your most efficient player.

For every Chris Andersen (126 rating, 11% usg in 2010) there's a JR Smith (101 rating 27% usg).  This does not mean the league is insane.  Chris would probably be even worse than 101 if he had to take 27% of the team's offense.  There are worse players out there than JR for his role, which includes shooting a bunch to cover for the bench's inability to create offense elsewhere.

Usage doesn't capture everything about role and the sharing of possessions.  Assists are probably the best way to capture moments where credit for scoring should be shared but usage doesn't consider them.  Since a Nugget got an assist on 70% of Andersen's buckets, his actual contribution to the offense is likely less than 11%!  Some portion of that usage is actually other players cooperating and splitting the workload with him in a way that he doesn't contribute back.

There are even more stats that are tracked (if not in the box score) that capture moments of teamwork well enough to say credit should be split.  Private geeks out there have numbers on ball-screens set, screens used, and post entry-passes that precede attempts.  Since these actions capture both the skill and decision-making of two teammates, they are like little assists that go unrecorded. Good screeners and entry passers improve their team's and teammates' statistics but not their own.

In a perfect world this score sheet data would be available to the causal geek, so we could look at events within a possession as well as events that end possessions.  Offense is about more than just who touches the ball last, which is all that usage tells you.  Between usage and assist data we can get a decent idea of a guy's role and how much credit or blame he should get for the final product, but this is just the beginning of the story.  Assignment is my name for everything else to keep in mind when looking at the advanced stats, because not everything that is a part of role is recorded just yet.