#HoopIdea: Treat Possessions Like They Matter!

Brook Lopez Free Throw
Why don’t we treat this like it matters?
Brook Lopez Free Throw
Why don’t we treat this like it matters?

ESPN’s HoopIdea was founded last season with the spirit of making basketball, already the best sport in the world, even better. But just as important as making it better, is understanding it.

Recently, I begun taking efficiency numbers from, as opposed to Basketball-Reference, because the two sites calculate possessions slightly differently, and I was more comfortable using the direct NBA numbers. But yet, something felt off.

As a self-proclaimed stathead, it’s easy to get wrapped up in successful formulas. Indeed, many of them are, though imperfect, excellent at helping understand a bigger or alternate picture. From the simple numbers like effective field goal percentage, which take into account the value of a three-point shot, to John Hollinger’s PER, which could use its own book on its much-inclusive calculation, using them in context helps you open a world of insight on the game of basketball.

But I’ve wondered something for a while now, exacerbated again by a conversation on Twitter today: why in the world do we estimate possessions?

PER needs to be calculated. Possessions don’t. Teams don’t have an uneven number of possessions in a game. A team doesn’t have 90.6 possessions in a game. They have 90, or 91.

This is how possessions are currently calculated, via Basketball-Reference’s Glossary:

0.5 * ((Tm FGA + 0.4 * Tm FTA - 1.07 * (Tm ORB / (Tm ORB + Opp DRB)) * (Tm FGA - Tm FG) + Tm TOV) + (Opp FGA + 0.4 * Opp FTA - 1.07 * (Opp ORB / (Opp ORB + Tm DRB)) * (Opp FGA - Opp FG) + Opp TOV))

The pieces are there, but the formula’s flaw is in its use of multipliers (multiplying free throws by .4, rebounds by 1.07) to calculate something tangible. It’s a starter’s kit for possession calculation, and like with most rough drafts, it’s a mess.

Possessions are a simple concept — it’s how many times your team had control of the ball in a position to score. And yet, we complicate it with excess multipliers. We don’t count possessions, we estimate them.

But why? There’s no need to estimate. There’s a set amount of possessions in every game!

Understanding possessions is crucial to understanding basketball, and the possession formula (in some iteration) is used in almost every reliable marker of efficiency: how much do you score (or allow), compared with how often you have the ball.

It’s dead simple, and yet, it’s wrong. It’s imperfect. It’s close, and it’s used as the estimate because it’s very good. But it’s still wrong, and it’s not “wrong” in an impossible-to-measure, theoretical PER sense, in which you’re attempting to assess value. It’s just a tangible, on-court number. And when figuring out the big concepts — who’s the most efficient player or team in the NBA, for instance — we’re relying on an archaic model when an even more archaic model — bean counting — beats it.

You could count possessions one of two ways. One is simple: have someone, at the game, count up how many possessions each team uses. Just count. Simple, easy, effective, done.

The second is a little more complicated, but much easier to automate: count up everything that ends a single possession, and subtract everything that maintains or extends a single possession.

In my estimation (no pun intended), that means you add:

  • Field goal attempts
  • Turnovers
  • Free throw attempts
  • end-of-quarters without a shot attempt

    Then subtract:

  • Offensive rebounds
  • Free throws from a technical foul that doesn’t change possession, from a flagrant foul, or from a defensive three seconds violation
  • # of fouls drawn that lead to free throws
  • # of fouls drawn on three-pointers
  • # of fouls drawn on and-ones

    (Note: I may be missing things. I, like the possession estimate, am quite imperfect.)

    Those last three are what get rid of the possession free throw multiplier: with fouls drawn that lead to free throws (on shooting fouls or in the bonus), you get two free throws, so taking one away equals one possession. Subtracting fouls on three-pointers takes away the third free throw. Fouls on and-ones are already part of a possession.

    The one qualm I still have with this is end-of-quarters without a shot attempt, because often teams don’t try to shoot, and shouldn’t count as a possession. For example, at the end of a blowout, a team will just run out the clock, or with .2 seconds at the half a team will merely inbound the ball and let the ticks run off. In that case, I can see having an arbiter, but that’s also a minor quibble.

    When assessing what teams and players are the best in the NBA in scoring or defending, we’ve done very well. We’ve gotten to a point in the so-called statistical revolution where we’re doing things better. But as former UNC coach and Michael Jordan disciple Dean Smith once said, basketball is a game of possessions.

    Now let’s treat them like they matter.