The Case of James Harden and the Challenge of Evaluating Net Rating Stats in the NBA

By: Bryan Povlinski on February 16, 2013

With the proliferation of advanced stats in the NBA the concept of Offensive, Defensive and Net Rating has grown in importance. Along with differential measures like +/- and improved shooting statistics like True Shooting % and Effective Field Goal % the concept of efficiency ratings has become well ingrained into the basketball stats culture. Offensive Rating and Defensive ratings were popularized by Dean Oliver in his book Basketball on Paper and are defined as:

Offensive Rating: Points produced per 100 possessions

Defensive Rating: Points allowed per 100 possessions

Net Rating: Offensive Rating – Defensive Rating

These ratings can be taken even further by calculating these ratings based on when a player is on the court vs. when he is off the court. Now you have On-Court/Off-Court Ratings that can help you understand how much a player is contributing to his team’s success. Theoretically, if a team has a higher Net Rating when a player is off the court than when he’s on-court then that player is hurting the team and the team would be better off with that player on the bench. However, that theory comes into question when I tell you that, according to this metric the 2012-2013 Houston Rockets are better off with James Harden on the bench.

Houston has a Net Rating of 1.6 when Harden is on the court and a rating of 3.8 when he’s on the bench. Take a look at the stats for Houston’s starting lineup from’s new stats tool at



From the standpoint of a fan that doesn’t make sense. The Rockets have been a big surprise this year after dumping almost their entire starting lineup from last year (Samuel Dalembert, Kyle Lowry, Luis Scola, Kevin Martin) and it seems like Harden has been the centerpiece. He’s been the leading scorer and the go-to guy in the clutch. How can the Rockets be a better team with James Harden on the bench? I took a look at some of his supporting stats throughout his career via

There doesn’t seem to be much of a drop-off from last season for Harden. His turnovers are up overall, but as a % they are exactly the same as last year. His shooting percentages are down slightly, but they’re still well above league average (especially for a shooting guard with as high of a usage rate as Harden). If you look at his On-Court vs. Off-Court splits last season with the Oklahoma City Thunder, Harden was one of the best on the team. He had a better on-court vs. off-court split than both Kevin Durant and Russell Westbrook.


As you can see, Harden was better on-court vs. off-court than both Westbrook and Durant, and was only slightly behind Nick Collison for the team lead. Is the difference that Houston has better back-ups for Harden than Oklahoma City did last season? The only true reserve at the SG and SF position for Houston is Carlos Delfino. His stats this season have been good, but I find it hard to believe that if he was getting the minutes James Harden is currently getting for the Rockets that they would playing anywhere near a playoff caliber level.

The only explanation I can get to is that Houston’s reserves have been extraordinarily good in garbage time. Whether they’re up by enough or down by enough to play all-reserve lineups for the entire 4th quarter the Rockets have been very good in those situations. Using a tool I’ve developed to look at specific lineup combinations it appears that my hypothesis is correct. Assuming that we can define garbage time for the Rockets as any minutes that did not include either James Harden or Chandler Parsons (those 2 play by far the most minutes and Delfino is the only legitimate backup to either player) the Rockets performance has been very good.



The way to read this graphic is that the Rockets have played 214.1 minutes with the combo of Harden and Parsons both off the court (those 2 being off court is the criteria for on-court in the tool). During those 214.1 minutes they’ve outscored opponents by 56 points, which, if extrapolated by a per-minute basis out to a 48 minute game the Rockets would outscore opponents by 12.55 points per game. With either Harden or Parsons on the court the Rockets outscore opponents by 1.97 points per game so the net difference would be 10.59.

I’m encouraged by’s release of all of their historical statistical information to possibly develop better metrics because if the metrics we’re currently using to evaluate players can be skewed this heavily by garbage time stats then there’s a lot of room for improvement.