APBRmetrics

Hey all,

over the past several years, the statistic of scoring - also known as points per game - has been completely decimated. It used to be that if someone scored 20 points per game, they were viewed as a top tier player and that was that. With a greater understanding of efficiency, nobody really cares about scoring totals anymore. One stat that hasn't really had a change in perception is rebounding.

People still view a good rebounder as someone who can get around 10 rebounds per game. However, rebounding numbers are very iffy. There are several issues such as:
team's shooting percentage and shot selection - fellow teammates ability to rebound etc.

This got me thinking of a better way to understand rebounding. At first, I started thinking about using the sportsvu which presents contested rebounding numbers.

While I think these are very important, they neglect the positioning aspect of rebounding. Part of rebounding is knowing where to be/go to actually have a chance at receiving a rebound which is completely ignored.

The next line of thinking which I've seen on this forum is measuring the impact a player has on their teams rebounding percentage. I think this is the best measure we have as of now and provides a fairly accurate picture. However, it fails to take into account two important things; the skill of the teammates and the back up.

Essentially, we have the same problems that come up in +/ compared to adjusted+/. If Dwight Howard is only be taking off for Omer Asik, then the changes in team rebounding percentage wont be meaningful. Similarly, if a team has 4 other amazing rebounders, one players impact will be less significant.

This leads me to wonder if there could be a stat similar to RAPM but only deals with rebounding? (measuring the impact one player has on their teams rebounding percentage adjusted for how good the other rebounders are on the court).
It seems as if it would provide the best measure of how a player impacts their teams ability to rebound, which is what we're really looking for here.

In the alternative, does anyone have other ideas or suggestions for the best ways to measure rebounding impact?

Crow wrote:http://stats-for-the-nba.appspot.com/ra ... _2014.html

as far as I know, this is not adjusted...

it only measures the impact on a team's rebounding percentage without taking into account the players on the floor, right?

ampersand5 wrote:

Crow wrote:http://stats-for-the-nba.appspot.com/ra ... _2014.html

as far as I know, this is not adjusted...

it only measures the impact on a team's rebounding percentage without taking into account the players on the floor, right?

It's adjusted, i.e. takes into account who's on the floor with you

Hopefully Jerry will respond. The title says adjusted rebounding. I think the offensive rebounding is. On defensive rebounding, I not sure the right way to use this data. Maybe IND def. reb. total - teammate impact?

JE once had straightforward adjusted rebounding factors.

JE - can you repost your methodology as the previous thread has been deleted.

It appears as if measuring a players impact on a team's rebounding totals taking into account the other players on the floor would be by far the best way to measure rebounding ability. To such an extent, I really wish we could have numbers for this season - and updated on a consistent basis, as it really needs to be something with more widespread attention.

Missed JE's almost simultaneous post above.

There are several old threads with info:
search.php?keywords=adjusted+rebounding ... mit=Search

See especially this one for explanation from Jerry.
viewtopic.php?f=2&t=8546&hilit=adjusted+rebounding

Thanks Crow!
I had read that thread previously - I was just looking to confirm that the same RAPM analysis was used as the factor to determine the data which I subsequently found J.E. confirming in another thread.

Anyways, going back to the OP - im shocked that nobody is really discussing this metric.

This seems to be the holygrail of rebounding statistics. I cannot think of any major flaws, and it seems futile to use any other rebounding metric as a result of this.

all hail J.E. I guess - I hope to get this metric more publicity if I can.
(and hopefully get some more current/sortable numbers)

Thanks for the kind words. I'll update with new data in a couple of days

I cannot think of any major flaws

I wouldn't call it a flaw, but this analysis can easily be improved by creating bayesian priors before running the RAPM(-type) regression, e.g. by regressing BoxScore stats onto long-term 'Adjusted Rebounding'

J.E. wrote:Thanks for the kind words. I'll update with new data in a couple of days

I cannot think of any major flaws

I wouldn't call it a flaw, but this analysis can easily be improved by creating bayesian priors before running the RAPM(-type) regression, e.g. by regressing BoxScore stats onto long-term 'Adjusted Rebounding'

J.E., are you going to run an actual RAPM like stat for rebounding like you did in the 2008-2011 study? I would love to see that. In the recent updates for rebounding, I believe you didn't include that. Only the players influence on his teammates and his rebounds per 100

Mentioned thread to Benjamin Morris on twitter...

JE - before I write something on your metric, are you going to be releasing "Estimating a players' influence on his teammates' BoxScore statistics using a modified Adjusted Plus Minus framework" before you present in a few weeks?

ampersand5 wrote:JE - before I write something on your metric, are you going to be releasing "Estimating a players' influence on his teammates' BoxScore statistics using a modified Adjusted Plus Minus framework" before you present in a few weeks?

You mean as a paper or sth? There's a very good chance I won't go to that conference, unfortunately

I just wanted to read the paper to get a better understanding of the methodlogy

Offensive rebounding and defensive rebounding are so different that lumping them together isn't all that useful - the biggest determinant of rebounding is the player's position when the shot goes up, and the ability to control that on offense is much greater than it is on defense.