Prior Event Adjusted Ratings

Apr 27, 2015; Portland, OR, USA; Portland Trail Blazers guard Damian Lillard (0) drives past Memphis Grizzlies forward Tony Allen (9) during the second quarter in game four of the first round of the NBA Playoffs at the Moda Center. Mandatory Credit: Craig Mitchelldyer-USA TODAY Sports

It’s common knowledge in basketball analytics that the best way to look at a team’s offense and defense is on a per possession basis; looking at how teams score and defend in the equivalent number of possessions. Normalizing the number of possessions teams get per game gives a more accurate number of how a team really scores.

But there’s something that per possession analysis doesn’t correct for: not all possessions are created equal. The way a possession starts has a huge impact on what happens during the possession. Here are average effective field goal percentage (eFG%) numbers for the possession following a certain outcome on the other end (via nbawowy):

So, after a dead-ball turnover, when the defense often has time to set, teams have had an eFG% of 46.5 this season. And after a live-ball turnover, which often leads to fast breaks, teams have had an eFG% of 61.5.

How the prior play ends has a huge effect on a team’s offensive and defensive ratings, which isn’t accounted for. Our perception of the Houston Rockets’ defense is distorted, because of their offense’s frequent liveball turnovers. It’s the same on the other end: the Wizard’s offense is at an inherent disadvantage because of the frequency at which their defense forces dead-ball turnovers. That’s not the Washington offense’s fault, but it still depresses their offensive rating and, in turn, our perception of their offensive ability.

Because of this effect, how we measure offense and defense isn’t entirely accurate.

Let’s look at the Portland Trailblazers. They sport one of the best offenses in the league scoring over 111 points per 100 possessions. What’s kept them from true contender status has been defense, as they rank 19th in the league in defensive efficiency. Offensive and defensive efficiency do a good job of telling us how good teams are overall, but don’t entirely split apart the two ends of the floor. This is especially true for the Blazers, where the gap between their offense and defense is overblown.

Portland’s offense commits dead ball turnovers at the fourth lowest rate in the league, so the Blazers’ defense starts very few defensive possessions set. Their defense also allows a far worse than average eFG% after made baskets –probably related to the Blazers being one of the best offensive rebounding teams in the league — and the Blazers’ offense makes an unusually high number of baskets. Just setting Portland’s prior event opportunity rates to league average rates drops their opponent eFG% by almost 1%.

It’s a different story on the offensive end. A high number of Blazers’ offensive possessions start with a defensive rebound — they don’t force many turnovers, of either variety–the second most efficient prior event. Adjusting to give them an average number of each prior event drops Portland’s offensive eFG% by over 1%.

I adjusted each team’s offensive and defensive eFG% to give them the league average number of each possession starter. This “true” eFG% should give a measure of a team’s shooting ability, independent of what happened on the other end of the floor.

I then ran a regression of the basic four factors* on offensive and defensive rating — the r-squared coming out to 0.97 — then used the provided formula to estimate offensive and defensive rating, subbing in my adjusted eFG% for standard eFG%.

*Free throw rate, rebound rate, and turnover rate likely change based on the prior events too, but right now there isn’t data available to quantify how. I suspect though, that some of the changes cancel each other out. For example, on fastbreaks following live ball turnovers, there are likely more fouls and fewer offensive rebounds.

Sortable chart

Now, this is in no way a more accurate measure of team strength. How a team ends its possessions matters a lot, as is illustrated above by some of the big changes in some team’s offensive and defensive ratings when set to an average distribution of prior events. That stuff matters when it comes to team strength, but offensive or defensive performance and offensive or defensive skill are two different things, the former being impacted by unrelated preceding factors. These ratings should provide a more nuanced understanding of true offensive and defensive ability, by isolating each from what terminates possessions on the opposite side of the floor.