RealGM

...would it make a meaningful difference in any stats/metrics? If so, which stats would look different?

Branched off from another thread. Original post:

This is where I smell some semantics, the old matter of whether an offensive rebound extends one possession or creates a new one. Yes, technically it's the former. I know. But it feels like it's effectively, essentially, all-but-technically the latter. I'm sure that there are dozens of threads about this on the Comparisons board already, and that you guys are sick of explaining it to n00bs and fogeys, and hearing rebuttals with terrible logic, etc. But bear with me. I apologize in advance. I'm a little right-brained.

There may be a statistical need to divide the game neatly in two, between when one team has the ball and when the other team has the ball, a need for it to be one or the other at any and every juncture in the game. But when playing, it certainly feels like there is a limbo region where neither team has possession of the ball, and that's when shots miss. It feels like every rebound is a kind of jump ball. Obviously, it's not literally a jump ball, which explicitly tries to create a 50/50 situation, carefully controlled by the ref, with just a touch of randomness. A few factors let teams/players increase or decrease their chances of snagging a rebound -- individual boxing out, team boxing out, anticipation of the ball's trajectory, taking/forcing good/bad shots, etc. But it does feel like the ball is neither team's when it's in the air.

So, not saying that this approach would be preferable, viable, rational, whatever...but humor me: What would the current stats look like if you did count offensive rebounds as new possessions? How would that affect overall efficiency, offensive efficiency, defensive efficiency? Maybe it wouldn't?

(p.s. I'm sure I'm using a bit of sloppy logic and misplaced vocab somewhere. But if you can get my gist, then I thank you.)

Warning: Sorry if my writing's a little jumbled, I'm writing this stream of conscious with no checking or editing.

Your new possession stat would essentially be what we currently call pace plus the number of offensive rebounds that game. Let's call your new stat pace* then and if we use this to replace our current definition of possession...

(Just going off the top of my head without checking the specific formulas, so someone please correct me if I'm wrong)

PER would be the same, WS would be different. Raw +/- would be the same, (R)APM would in theory be different since you're weighting by the # of possesions, but what matters there is the ratio of possesions played by one lineup versus another so in practice if you increase all possessions by the same percentage things would be unchanged.

If we want to measure how often teams run up and down the court, pace* would be a poor measurement. Pace* would rate plodding half court teams with strong o-rebounding like the 2009 Blazers faster than they were relative to the rest of the league.

Points per possession wouldn't be a good measure of overall efficiency any more. Consider this: would you rather make a layup, or miss a layup, get the offensive rebound, then make a 3? Obviously you would always prefer the second, but the new points per possession tells you that the former is better. TS% gets this wrong also, but no one knowledgeable uses TS% as a measure of overall offensive efficiency, only shooting efficiency.

In summary - disadvantages: new definition no longer measures pace, or overall efficiency well
advantages: ???? can't think of any off the top of my head

Going through this quick, so let me know if there's a detail you covered that I missed:

-I believe that technically in accordance to the rule book, the wording implies that every rebounding starts a new possession.

-When statisticians started getting serious though, they either ignored this, or didn't check it at all, because Offensive and Defensive Efficiency stats are far more straight forward to make meaningful use of if you don't call an offensive rebound the start of the new possession.

When an offense gets the ball and brings it across the half court line, their goal is essentially just to score as many points as possible before relinquishing the ball (obviously there's a touch more to it than that). To that end, the offensive strategy is developed not simply to try to get one good shot off, but also when possible to get the ball back after that shot for another crack at it. Hence, Dennis Rodman has a valuable role to play in helping his team's offense be as successful as possible even if he isn't there to score himself (and hence, Michael Jordan can claim to have played on exceptionally good offenses).

One might argue that it would be more accurate if Rodman's contributions here weren't included as "offense". I'm not going to say that's wrong. However, it would be very confusing when people tried to do apples-to-apples comparisons of offenses if we ignored the role guys like Rodman were playing to support their teammates while their teammates were playing "offense".

Of course, as is, there are still difficulties when using an all purpose Offensive Rating. The Celtics prioritize getting back on transition defense ahead of crashing the offensive boards, so quite literally, the team is putting less offensive effort in than other teams, and an Offensive Rating won't tell you why, and might imply a general incompetence rather than a strategic choice. But that's why we have more detailed stats to go along with the broad strokes ones.

Doctor MJ wrote:Going through this quick, so let me know if there's a detail you covered that I missed:

-I believe that technically in accordance to the rule book, the wording implies that every rebounding starts a new possession.

-When statisticians started getting serious though, they either ignored this, or didn't check it at all, because Offensive and Defensive Efficiency stats are far more straight forward to make meaningful use of if you don't call an offensive rebound the start of the new possession.

Having participated in the conversation that ultimately settled on the "offensive rebounds don't create new possessions" "rule", the perspective that every rebound creates a new possession was not ignored. We analyzed it to death. The analysis was more robust when going with "offensive rebounds extend possessions" than it was otherwise. "Efficiency" actually meant something rather than just being an interesting stat.

I'm not sure whether the official rule book was consulted, but it wouldn't have persuaded us one way or another. The rule book is about governing/running a game, not about generating meaningful statistical data. Possession rules of the game are for letting refs know when to reset the shot clock.

When an offense gets the ball and brings it across the half court line, their goal is essentially just to score as many points as possible before relinquishing the ball (obviously there's a touch more to it than that). To that end, the offensive strategy is developed not simply to try to get one good shot off, but also when possible to get the ball back after that shot for another crack at it. Hence, Dennis Rodman has a valuable role to play in helping his team's offense be as successful as possible even if he isn't there to score himself (and hence, Michael Jordan can claim to have played on exceptionally good offenses).

One might argue that it would be more accurate if Rodman's contributions here weren't included as "offense". I'm not going to say that's wrong. However, it would be very confusing when people tried to do apples-to-apples comparisons of offenses if we ignored the role guys like Rodman were playing to support their teammates while their teammates were playing "offense".

As DeanO once said, rebounding is not a separate part of the game. It's not like football where there's offense, defense and special teams. In basketball, there's offense and defense. Offensive rebounding is part of offense; defensive rebounding is part of defense.

Of course, as is, there are still difficulties when using an all purpose Offensive Rating. The Celtics prioritize getting back on transition defense ahead of crashing the offensive boards, so quite literally, the team is putting less offensive effort in than other teams, and an Offensive Rating won't tell you why, and might imply a general incompetence rather than a strategic choice. But that's why we have more detailed stats to go along with the broad strokes ones.

I wouldn't consider this example to be a difficulty with offensive rating. Teams make strategic trade-offs all the time. Boston, for example, sacrifices some offensive efficiency (through getting offensive rebounds) in exchange for getting back on defense and limiting the opponent's chances of an easy score. It would be reasonably simple to do a "back of the spreadsheet" calculation to see if that strategic choice makes sense. Knowing some of their stat guys (like Mike Zarren), I'm sure they've done the math and could tell you exactly what they're gaining on the defensive end versus what they're "giving up" on offense.

Nivek wrote:As DeanO once said, rebounding is not a separate part of the game. It's not like football where there's offense, defense and special teams. In basketball, there's offense and defense. Offensive rebounding is part of offense; defensive rebounding is part of defense.

I would disagree. I think there are basically three aspects of a game.

1) How well teams play on offense
2) How well teams play on defense
3) How well teams control possessions.

I think #3 is overlooked all the time, and it's a shame. Possession control impacts the game just as much as the other two, and we miss a lot of insight by merging it with either offense or defense.

A guy like Rodman was absolutely dominant in this regard. Think of how many extra possesions he got his team, and how that effected the final score. That nuance shouldn't be merged with off/def, but analyzed on it's own merits.

1995 Bulls:
for/opp
41.5 rpg/40.4 rpg +1.1
81.8 fga/78.0 fga +3.8

1996 Bulls
44.6 rpg/38.0 rpg +6.6
84.0 fga/78.4 fga +5.6
^
With Rodman, we can see that he had a rather significant impact on the ratio of possessions Chicago got. Whether they scored or not should be all that's reflected on offensive eff, not whether Rodman bailed them out with an off. rebound.

I'm working on a project that will clarify this a bit more in the coming weeks.

Nivek wrote:It would be reasonably simple to do a "back of the spreadsheet" calculation to see if that strategic choice makes sense.

Just run a regression on the data. The result will show you that ORB% has no positive impact on the game. A correlation analysis will show that the better teams have in average a lower ORB% than the worse teams.

The concept is also pretty simple to understand. Due to the defensive setting of a team it is more likely that the defending team is getting the rebound than the offensive team. You can make that a bit more sophisticated, if you explicitly look at where the ball ends up on the court after a missed shot.

It is also pretty reasonable sending the bigger guys back on defense rather than let them crash the boards, because with the offensive setting in which a guy like Garnett is rather out on the perimeter than close to the basket, it makes much more sense to have him improving the transition defense than let him try to work from the perimeter to get the offensive board.

An Unbiased Fan wrote:

Nivek wrote:As DeanO once said, rebounding is not a separate part of the game. It's not like football where there's offense, defense and special teams. In basketball, there's offense and defense. Offensive rebounding is part of offense; defensive rebounding is part of defense.

I would disagree. I think there are basically three aspects of a game.

1) How well teams play on offense
2) How well teams play on defense
3) How well teams control possessions.

I think #3 is overlooked all the time, and it's a shame. Possession control impacts the game just as much as the other two, and we miss a lot of insight by merging it with either offense or defense.

Except, of course, that possession control is explicitly included in DeanO's four factors, which are the basis for offensive and defensive ratings, which is what people are talking about when discussing efficiency.

A guy like Rodman was absolutely dominant in this regard. Think of how many extra possesions he got his team, and how that effected the final score. That nuance shouldn't be merged with off/def, but analyzed on it's own merits.

1995 Bulls:
for/opp
41.5 rpg/40.4 rpg +1.1
81.8 fga/78.0 fga +3.8

1996 Bulls
44.6 rpg/38.0 rpg +6.6
84.0 fga/78.4 fga +5.6
^
With Rodman, we can see that he had a rather significant impact on the ratio of possessions Chicago got. Whether they scored or not should be all that's reflected on offensive eff, not whether Rodman bailed them out with an off. rebound.

Rodman's rebounds are part of the team's offensive rating and his individual offensive rating. He didn't get his team "extra possessions," he got his team extra opportunities to score on their possessions. Those extra opportunities were extremely valuable to his team.

Offensive efficiency at both the team and individual levels includes the value of offensive rebounding.

mysticbb wrote:

Nivek wrote:It would be reasonably simple to do a "back of the spreadsheet" calculation to see if that strategic choice makes sense.

Just run a regression on the data. The result will show you that ORB% has no positive impact on the game. A correlation analysis will show that the better teams have in average a lower ORB% than the worse teams.

Are there stat goobers around who haven't run this regression?

I think this is one of the first things I "discovered" when I started looking at NBA stats back in the late-1980s. I didn't know what a regression was at the time, but I noticed that the "good" offensive rebounding teams tended to not be good teams.

The concept is also pretty simple to understand. Due to the defensive setting of a team it is more likely that the defending team is getting the rebound than the offensive team. You can make that a bit more sophisticated, if you explicitly look at where the ball ends up on the court after a missed shot.

It is also pretty reasonable sending the bigger guys back on defense rather than let them crash the boards, because with the offensive setting in which a guy like Garnett is rather out on the perimeter than close to the basket, it makes much more sense to have him improving the transition defense than let him try to work from the perimeter to get the offensive board.

Good points. What I'm saying, though is that with the information we have available now, it'd be pretty easy to estimate how many points of offensive efficiency are lost due to largely ignoring offensive rebounding vs. how many points of defensive efficiency are gained due to getting back on defense instead of hitting the offensive boards. I don't have time to do it at the moment, but it's possible.

Consider this: would you rather make a layup, or miss a layup, get the offensive rebound, then make a 3?

Kabstah, I'm sure the numbers have already been crunched to death, but...what percentage of 1) missed layups are 2) rebounded by the offense and 3) wind up turning into a 3? If it's a really low percentage, then I'd rather just make the layup. Mathematically, though, if I had control over the outcome of every play, I'd obviously rather miss the layup and get a 3 every time up the court.

Having participated in the conversation that ultimately settled on the "offensive rebounds don't create new possessions" "rule", the perspective that every rebound creates a new possession was not ignored. We analyzed it to death. The analysis was more robust when going with "offensive rebounds extend possessions" than it was otherwise. "Efficiency" actually meant something rather than just being an interesting stat.

I'm not sure whether the official rule book was consulted, but it wouldn't have persuaded us one way or another. The rule book is about governing/running a game, not about generating meaningful statistical data. Possession rules of the game are for letting refs know when to reset the shot clock.

Nivek, this is really interesting. So, even if an offensive rebound (and all rebounds) were in reality a kind of third species of possession, a variable jump ball type of thing, you would choose to ignore that on-the-court reality in order to produce a meaningful measure. That seems strange, but I think I can understand it. Sometimes in order to capture the overall reality of the game in numbers, you have to ignore/manipulate certain realities of the game? It's like a pictoral map versus a painfully precise topographical one, maybe?