Bounding and Stealin’: Examining the Value of Rebound Types

Jan 22, 2014; Phoenix, AZ, USA; Indiana Pacers forward Paul George (left) and guard Lance Stephenson against the Phoenix Suns at US Airways Center. Mandatory Credit: Mark J. Rebilas-USA TODAY Sports

As the Pacers imploded over the final weeks of the 2013-14 season, one possible cause advanced for the meltdown was internal friction. One of the principal issues was said to be Lance Stephenson’s penchant for “stealing” rebounds from teammates. Similarly, some of Kevin Love’s defensive shortcomings are often attributed to his desire to chase every possible carom to add to his own stats.

The notion of stealing rebounds and padding ones own stats is one of those things that is generally known to be happening, but not really fully understood. Casual empiricism while watching NBA basketball will show that a large number, perhaps even a majority, of missed shots are controlled rather effortlessly by the defense. These rebounds ending “one and done” possessions aren’t much of a reflection of the skill of the player who secures them. For the most part, if one defender didn’t casually scoop up the ball, another would have. Rebound thieves are players who seem to have an unseemly interest in chasing these “free” defensive rebounds.

With SportVU’s ability to dive deeper into rebounding, a more thorough inquest of this phenomenon is possible, though in a much less judgy manner. First the results, among players who averaged 5 or more rebounds per game last season, those whose rebounds were “worth” less and more than indicated by simple RPG, respectively:

Without belaboring methodology, this analysis posits defensive rebounding % (DREB%) is a function of:

• A team’s ability to secure those “uncontested” rebounds discussed above;
• The team’s ability to prevent the opposition from grabbing uncontested offensive rebounds[1. Or, in other words to contest as many defensive rebounds as possible.]; and
• To then win those contested rebound battles.

In a simple regression model, all three were found to be highly significantly correlated with DREB%. In that model, each uncontested defensive rebound was “worth” about half as much as other rebounds. This makes some logical sense, as there is clearly some value to securing these free rebounds in that having players attack rebounds (and thus prevent free opposing offensive rebounds is a good thing. The value is in the pursuit as much as securing the rebound itself, and since contested rebound battles are almost evenly split between offensive and defensive wins, the pursuit is worth that “half” rebound. After the value of uncontested defensive rebounds was halved, the total for all players was scaled to equal the total rebounds available, simply to present these “adjusted” rebounds on the same scale as the more familiar rebounds per game.

As the above chart shows, many of the players who may not have been adding as much value with their rebounding are peremetire typers who secure freebie defensive boards and little more, while those who benefited most were the bangers who not only hit the offensive glass, but fought and won battles on the defensive backboards. On the negative side, Stephenson does make a prominent appearance, but he wasn’t even the worst offender among Pacers, with Paul George taking that honor. On the other side of the ledger, Roy Hibbert appears to have been the victim of this thievery, perhaps aligning his play more to the needs of the team than his own numbers.

Kevin Love? Last season his adjusted rebounds were actually about .25 per game higher than his actual rebounds, and that was in a season in which he posted a career low offensive rebound percentage. His continued move to the outside in Cleveland this season has seen a much higher proportion of his rebounds coming in the less valuable “uncontested defensive” form.

The above is all very preliminary, there are any number of improvements and caveats to be made. Certainly, a more robust methodology using the full data set[2. Such as Second Spectrum’s “Dimensions of Rebounding” paper from the 2014 Sloan Conference.] could provide more precision both in terms of inputs[3. As with many public SportVU numbers, “contested” rebounds are precisely defined in a way which might not precisely align with common use of the term, and the count probably both under and overcounts rebounding attempts based on circumstances.] and valuing outputs more accurately. Similarly, the discount applied to uncontested defensive rebounds does not account for such theories as theories stating that defensive rebounds by point guards can lead to better ensuing offensive possessions.[4. If memory serves, there was an old APBR thread about this effect with respect to Jason Kidd specifically.] Finally, there is the ever present problem which is especially acute in rebounding in terms of assigning the value from a team accomplishment to one individual player.

Still while there are some known issues, this represents a decent first pass at better valuing various types of rebounds.