RealGM

Hey y'all,

With the way teams have decided not to compete for offensive rebounds, I've been feeling like I'd really like to have the best method possible for ballparking the efficacy of offenses given what they "put in". Basically, a 3/4th 4 factor measure trying to normalize for the effect of offensive rebounding.

(In particular I've been thinking about the current Miami squad because they seem so extreme in this.)

To do this right, you'd have to put a lot of though and a lot of number crunching into it. Obviously no solution would be perfect, but since ORtg is itself seeming to be increasingly non-apples-to-apples from team to team, this additional metric would be nice to see to.

I feel like someone has to have already done this, but I don't know who or where. I want to see what they've done.

Out of my own curiosity though I made my "slap dash" method. (You don't have to tell me it's not good enough - I know that, but it's what I could gin up quickly).

Basically take the TS% method, apply it to the whole team without the "make it look like FG%" adjustment, then treat TOs like an additional shot. Multiply it all by 100. Simplistic thinking: In the absence of ORs, missed shots + turnovers basically are the number of possessions.

Top teams I see on this metric along with their rating:

1. '08 Phoenix 102.2
2. '07 Phoenix 102.1
3. '14 Miami 102.0
4. '13 Miami 101.4
5. '10 Phoenix 101.2
6. '88 Boston 101.0
7. '95 Utah 100.1
8. '09 Phoenix 100.0
9. '05 Phoenix 100.0
10. '06 Phoenix 99.8

Other noteworthies (each "team" only listed once.

11. '11 Denver 99.7
12. '13 Oklahoma City 99.4
14. '08 Orlando 99.3 (Might surprise folks, but they didn't do ORs really)
15. '87 Lakers 99.3
16. '08 Lakers 99.2

Skipping to teams of particular interest:
31. '92 Chicago 98.0 (Best Jordan Bulls team - great ORtg, but with huge ORs)
98. '79 Lakers 93.7 (Best pre-80s team)

Doctor MJ wrote:Hey y'all,

With the way teams have decided not to compete for offensive rebounds, I've been feeling like I'd really like to have the best method possible for ballparking the efficacy of offenses given what they "put in". Basically, a 3/4th 4 factor measure trying to normalize for the effect of offensive rebounding.

(In particular I've been thinking about the current Miami squad because they seem so extreme in this.)

To do this right, you'd have to put a lot of though and a lot of number crunching into it. Obviously no solution would be perfect, but since ORtg is itself seeming to be increasingly non-apples-to-apples from team to team, this additional metric would be nice to see to.

I feel like someone has to have already done this, but I don't know who or where. I want to see what they've done.

Out of my own curiosity though I made my "slap dash" method. (You don't have to tell me it's not good enough - I know that, but it's what I could gin up quickly).

Basically take the TS% method, apply it to the whole team without the "make it look like FG%" adjustment, then treat TOs like an additional shot. Multiply it all by 100. Simplistic thinking: In the absence of ORs, missed shots + turnovers basically are the number of possessions.

Top teams I see on this metric along with their rating:

1. '08 Phoenix 102.2
2. '07 Phoenix 102.1
3. '14 Miami 102.0
4. '13 Miami 101.4
5. '10 Phoenix 101.2
6. '88 Boston 101.0
7. '95 Utah 100.1
8. '09 Phoenix 100.0
9. '05 Phoenix 100.0
10. '06 Phoenix 99.8

Other noteworthies (each "team" only listed once.

11. '11 Denver 99.7
12. '13 Oklahoma City 99.4
14. '08 Orlando 99.3 (Might surprise folks, but they didn't do ORs really)
15. '87 Lakers 99.3
16. '08 Lakers 99.2

Skipping to teams of particular interest:
31. '92 Chicago 98.0 (Best Jordan Bulls team - great ORtg, but with huge ORs)
98. '79 Lakers 93.7 (Best pre-80s team)

Interesting stuff Doc, just had a question or 2. In using the shooting percentages, missing the first of 2 FTs wouldn't affect this correct? As well, I was just checking the formula for a possession, was this basically that except removing anywhere that O Rebounds were? That would be:

0.5 * (Tm FGA - Tm FG) + Tm TOV) + (Opp FGA + 0.4 * Opp FTA ) * (Opp FGA - Opp FG) + Opp TOV))

bondom34 wrote:Interesting stuff Doc, just had a question or 2. In using the shooting percentages, missing the first of 2 FTs wouldn't affect this correct? As well, I was just checking the formula for a possession, was this basically that except removing anywhere that O Rebounds were? That would be:

0.5 * (Tm FGA - Tm FG) + Tm TOV) + (Opp FGA + 0.4 * Opp FTA ) * (Opp FGA - Opp FG) + Opp TOV))

Okay let me elaborate - but I'll say up front my equation is much simpler than that.

In real basketball, these are the events whose existences necessitate the end of a possession:

A made Field Goal.
A missed Field Goal Attempt which is not rebounded by the offense.
Certain Free Throws.
A turnover.

In this simplified view I"m using , the part in red above goes away. Now consider which box score stats would be needed to add up all these events:

FGA = A made Field Goal or a missed Field Goal Attempt
??? * FTA = Some fraction of Free Throws
TO = Turnover

We would then expect a Point / Posseession metric to take the number of points scored and divide by these factors.

SlapDash ORtg = Points / (FGA + ??? * FTA + TO)

Alright, now look at the equation for TS%:

TS% = Points / (FGA + 0.44*FTA) / 2

The "2" in TS% is just cosmetic. It's there to make it look more like FG%. So this means the only differences between the SlapDash metric and TS% are:

1) A mysterious number 0.44 instead of the ??? . Could that be the answer we're looking for? Pretty much. The 0.44 comes from an analysis of what fraction of FTAs actually constitute clinching an end to a possession. So - coming back to answer your question - front end FTAs would be among those which are in the "does not clinch" category. They are indeed factored in.

2) The SlapDash has the additional Turnover factor. Which it should.

It might seem weird that I'm taking a shooting efficiency method as the basis for an overall possession metric, but I don't see anything wrong with it. They aren't identical, but the difference is clear enough.

Now, last step, my cosmetic fix: ORtg is measured by 100 possessions instead of by 1, so I'll multiply by 100:

SlapDash = 100 * Points / (FGA + 0.44*FTA + TO)

Pretty simple. Too simple of course, but I don't think simple to the point of worthlessness if we really can't find a heavy duty metric from someone.

Doctor MJ wrote:

bondom34 wrote:Interesting stuff Doc, just had a question or 2. In using the shooting percentages, missing the first of 2 FTs wouldn't affect this correct? As well, I was just checking the formula for a possession, was this basically that except removing anywhere that O Rebounds were? That would be:

0.5 * (Tm FGA - Tm FG) + Tm TOV) + (Opp FGA + 0.4 * Opp FTA ) * (Opp FGA - Opp FG) + Opp TOV))

Okay let me elaborate - but I'll say up front my equation is much simpler than that.

In real basketball, these are the events whose existences necessitate the end of a possession:

A made Field Goal.
A missed Field Goal Attempt which is not rebounded by the offense.
Certain Free Throws.
A turnover.

In this simplified view I"m using , the part in red above goes away. Now consider which box score stats would be needed to add up all these events:

FGA = A made Field Goal or a missed Field Goal Attempt
??? * FTA = Some fraction of Free Throws
TO = Turnover

We would then expect a Point / Posseession metric to take the number of points scored and divide by these factors.

SlapDash ORtg = Points / (FGA + ??? * FTA + TO)

Alright, now look at the equation for TS%:

TS% = Points / (FGA + 0.44*FTA) / 2

The "2" in TS% is just cosmetic. It's there to make it look more like FG%. So this means the only differences between the SlapDash metric and TS% are:

1) A mysterious number 0.44 instead of the ??? . Could that be the answer we're looking for? Pretty much. The 0.44 comes from an analysis of what fraction of FTAs actually constitute clinching an end to a possession. So - coming back to answer your question - front end FTAs would be among those which are in the "does not clinch" category. They are indeed factored in.

2) The SlapDash has the additional Turnover factor. Which it should.

It might seem weird that I'm taking a shooting efficiency method as the basis for an overall possession metric, but I don't see anything wrong with it. They aren't identical, but the difference is clear enough.

Now, last step, my cosmetic fix: ORtg is measured by 100 possessions instead of by 1, so I'll multiply by 100:

SlapDash = 100 * Points / (FGA + 0.44*FTA + TO)

Pretty simple. Too simple of course, but I don't think simple to the point of worthlessness if we really can't find a heavy duty metric from someone.

Thanks for the clarification Doc! It seems unless you really do a play by play analysis, this is about the simplest way, and play by play would take a pretty good amount of time and/or manpower.

I guess I'm left wondering why you'd want to remove the O rebounding effect at all. I understand that it's more of an afterthought today, but teams still consider rebounding when putting together a lineup.

If I can get an extra average scorer out there at the expense of removing my subpar scoring, but excellent rebounding big, is my offense really better?

Personally, I favor cutting the 4 factors down to 3, but combining efg% and fta/fga into simple TS%.

Chicago76 wrote:I guess I'm left wondering why you'd want to remove the O rebounding effect at all. I understand that it's more of an afterthought today, but teams still consider rebounding when putting together a lineup.

If I can get an extra average scorer out there at the expense of removing my subpar scoring, but excellent rebounding big, is my offense really better?

I have to agree with this. I also think teams are undervaluing offensive rebounding.

I took a far more complicated swag at normalizing this a few years back. The goal was to control for a team's "get back on D" strategy. The method went like this:

1) look for individual outliers on the team, by position, in terms of OREB%

e.g. Calculate weighted-avg. of OREB% of a team's PF/Cs. If there's an outlier, remove him -- the thinking is his individual oreb% isn't determined by team strategy, but by his own skillful or renegade ways.

2) Do this for all teams, and determine an OREB% index (who crashes the boards the most and the least)

3) Then simply normalize ORTGs based on the league averages. So a team at 95% of the league average gets a small boost in ORTG which is comparable to the penalty a a team at 105% OREB would get.

I did this for a few years if anyone is interested.

I would also add that I do not like this for comparing across eras, because the strategies regarding rebounding have changed so much. And of course, if the goal is to normalize for team strategy, we're still left with the giant confound that team's with players who scramble the defense will naturally have higher OREB% regardless of strategy because those misses rebounded at a higher percentage by the offense in the natural course of the play.

ElGee wrote:I took a far more complicated swag at normalizing this a few years back. The goal was to control for a team's "get back on D" strategy. The method went like this:

1) look for individual outliers on the team, by position, in terms of OREB%

e.g. Calculate weighted-avg. of OREB% of a team's PF/Cs. If there's an outlier, remove him -- the thinking is his individual oreb% isn't determined by team strategy, but by his own skillful or renegade ways.

2) Do this for all teams, and determine an OREB% index (who crashes the boards the most and the least)

3) Then simply normalize ORTGs based on the league averages. So a team at 95% of the league average gets a small boost in ORTG which is comparable to the penalty a a team at 105% OREB would get.

I did this for a few years if anyone is interested.

I would also add that I do not like this for comparing across eras, because the strategies regarding rebounding have changed so much. And of course, if the goal is to normalize for team strategy, we're still left with the giant confound that team's with players who scramble the defense will naturally have higher OREB% regardless of strategy because those misses rebounded at a higher percentage by the offense in the natural course of the play.

Interesting. Yes, I would like to see what that data looks like if you can share it without much trouble.

Chicago76 wrote:I guess I'm left wondering why you'd want to remove the O rebounding effect at all. I understand that it's more of an afterthought today, but teams still consider rebounding when putting together a lineup.

If I can get an extra average scorer out there at the expense of removing my subpar scoring, but excellent rebounding big, is my offense really better?

Personally, I favor cutting the 4 factors down to 3, but combining efg% and fta/fga into simple TS%.

I want to normalize because right now the lack of O rebounding makes teams look like they are doing something wrong compared to standards of the past. If this isn't an error but a strategy, I'd like a metric that sees it as such.

This is funny because I did the same thing to highlight how terrorizing Miami's offense has been to start the season:
http://ascreamingcomesacrossthecourt.bl ... nning.html

I went back to the introduction of the turnovers. I too used PTS/(FGA+.44*FTA+TOV) because I just wanted to simply explain how unique Miami's start was. (They've cooled off since a bit.)

I'm not sure how complicated a formula you would really need here. You can't exactly calculate it from such a simple formula because there are a number of ways that can go wrong: possessions that end because the period ends, technical/flagrant foul free throws, and And-1's. I might look into this.

Small update on this.

2014 Top 10

MIA 102.2
LAC 98.5
DAL 98.2
SAS 98.1
HOU 96.9
POR 96.7
OKC 96.5
PHO 95.2
ATL 95.0
BKN 93.8

There needs to be some adjustment made for league environment/defenses played. Going by ATC's method, they're at +9.4 right now which would bef 3rd all time ahead of the 2013 Heat at +9.1.

ElGee wrote:I took a far more complicated swag at normalizing this a few years back. The goal was to control for a team's "get back on D" strategy. The method went like this:

1) look for individual outliers on the team, by position, in terms of OREB%

e.g. Calculate weighted-avg. of OREB% of a team's PF/Cs. If there's an outlier, remove him -- the thinking is his individual oreb% isn't determined by team strategy, but by his own skillful or renegade ways.

2) Do this for all teams, and determine an OREB% index (who crashes the boards the most and the least)

3) Then simply normalize ORTGs based on the league averages. So a team at 95% of the league average gets a small boost in ORTG which is comparable to the penalty a a team at 105% OREB would get.

I did this for a few years if anyone is interested.

I would also add that I do not like this for comparing across eras, because the strategies regarding rebounding have changed so much. And of course, if the goal is to normalize for team strategy, we're still left with the giant confound that team's with players who scramble the defense will naturally have higher OREB% regardless of strategy because those misses rebounded at a higher percentage by the offense in the natural course of the play.

Whenever you're around I'd be interested in seeing that

SideshowBob wrote:Small update on this.

2014 Top 10

MIA 102.2
LAC 98.5
DAL 98.2
SAS 98.1
HOU 96.9
POR 96.7
OKC 96.5
PHO 95.2
ATL 95.0
BKN 93.8

There needs to be some adjustment made for league environment/defenses played. Going by ATC's method, they're at +9.4 right now which would bef 3rd all time ahead of the 2013 Heat at +9.1.

Make that 102.5 (#1 Raw) and +9.6 (#2 adjusted) all time now.

Just remember if you're removing points from offensive rebounds out of ortg you have to include those points in drtg. You would do this by subtracting offensive rebounds points from total points allowed and then conducting the possession adjustment.

If you don't do that you are just conducting incomplete analysis. Frankly, I still think it is silly to remove offensive rebounds points from ortg. There is a far better case for crediting points scored off live ball turnovers to drtg rather than removing offensive rebounds from ortg

sp6r=underrated wrote:Just remember if you're removing points from offensive rebounds out of ortg you have to include those points in drtg. You would do this by subtracting offensive rebounds points from total points allowed and then conducting the possession adjustment.

If you don't do that you are just conducting incomplete analysis. Frankly, I still think it is silly to remove offensive rebounds points from ortg. There is a far better case for crediting points scored off live ball turnovers to drtg rather than removing offensive rebounds from ortg

The argument is not for a complete dismissal of offensive rebounding and its consequences on offense, it's that a separate window is needed to judge the successful of an offense in what it's attempting to do.

Doctor MJ wrote:

sp6r=underrated wrote:Just remember if you're removing points from offensive rebounds out of ortg you have to include those points in drtg. You would do this by subtracting offensive rebounds points from total points allowed and then conducting the possession adjustment.

If you don't do that you are just conducting incomplete analysis. Frankly, I still think it is silly to remove offensive rebounds points from ortg. There is a far better case for crediting points scored off live ball turnovers to drtg rather than removing offensive rebounds from ortg

The argument is not for a complete dismissal of offensive rebounding and its consequences on offense, it's that a separate window is needed to judge the successful of an offense in what it's attempting to do.

I understand your position. It is as follows:

Miami made a decision to deemphasize offensive rebounds to help their defense. This decision is part of a larger league trend in which teams have deemphasized offensive rebounds on behalf of protecting their defense. Accordingly, their offense is more effective than traditional ortg shows. Thus it makes sense to try to calculate ortg without offensive rebounds to determine their true offensive effectiveness. To not do so is to punish them for making what is a “clearly strategic choice.”

I think that is a fair summary. I’ve discussed this matter with you before so I’m not really interested in rehashing the debate with you. In our conversation, you acknowledged at the end that when you are doing is mostly shifting efficiency from the defense to the offense:

Doctor MJ wrote:Miami's not a great rebounding team, but offensive rebounding numbers make it look like a huge weakness when it's clearly a strategic choice to some degree.

And yes, to the extent that the rebounding effort is just shifted, what it means is "inflating" how good the defense looks while the offense gets "deflated". My defense of their offense then can be taken partially as a criticism of their defense...but it seems to go without saying that the criticism is a mild one.

viewtopic.php?f=64&t=1288806&start=30#p37757983

I haven’t had a chance to discuss it with sideshowbob. He’s a great poster so I’m sure he already knows this on some level but I want him to be explicit that when he is removing offensive rebounds from ortg he is merely shifting points off offensive rebounds to drtg. If you don’t do that you are merely waiving away the tradeoffs from strategic choices.

I think you mentioned in the past that you work in education so I’ll draw an analogy from that field. A student decides to devote a disproportionate amount of his study time to the SAT rather than towards maintaining his High School GPA. He believes that this will maximize his chances of getting into his school of choice. The strategy works his SAT is higher than it would normally have been but his GPA did dip a bit. If he were to argue that his GPA was deflated due to a strategic choice to enhance his SAT he is implicitly arguing that his SAT score is inflated. In school, as in basketball, as in life there are tradeoffs. The costs have to be included somewhere in the ledger.

I’ve meandered a bit but I’ll finish in this point. The difference between offense and defense is not nearly as black and white as people think. That is why I focus much more on overall point differential rather than ortg and drtg as separate items.

While the case for removing offensive rebounds from ortg is quite weak, IMO, there is a really strong argument for considering points off liveball turnovers to be part of drtg. The defense can credibly claim credit for points that occur when a player steals a pass setting his team up for a one on zero fastbreak. That said, when making that argument I fully recognize I am also saying their offense is not as good as their overall ortg says.
I want to make sure sideshowbob recognizes that is the consequence of his position for offensive rebounds.

sp6r, I think our previous conversations are coming back to me now. General position:

I understand the fear of essentially double counting. If I say this is a GOAT offense based on what they do without offensive rebounding, and then praise the back-on-transition-aided defense to the heavens, it sounds well and good, but in the end a team is what differential it produces relative to its opponent. There ain't no such thing as a free lunch, so skimping in one area is having a negative effect that should be factored in somewhere.

My interest here is as someone who likes to split things into factors as best I can. I like analyzing a team's offensive effectiveness. I know there are factors involved that make one perfect apples-to-apples comparison impossible, but I feel compelled to do as best I can. Adding a method like this one in addition to the full ORtg and other stuff gives me an additional window through which to draw insight from.

Doctor MJ wrote:My interest here is as someone who likes to split things into factors as best I can. I like analyzing a team's offensive effectiveness. I know there are factors involved that make one perfect apples-to-apples comparison impossible, but I feel compelled to do as best I can. Adding a method like this one in addition to the full ORtg and other stuff gives me an additional window through which to draw insight from.

That is a totally reasonable thing to do and has a lot of value.

Just wanted to bump this as it's the closest topic to what I was going to mention, and didn't want to start a new thread. Hickory High did an interesting write up of adjusting O/D Ratings based on how a possession begins (Rebound, dead/live ball TO, made basket), and adjusting for the average eFG for how the possesion starts.

http://www.hickory-high.com/the-differe ... and-skill/

Basically shows the effects of teams committing a large amount of live ball TOs, or creating dead ball TOs being at a slight disadvantage defensively/offensively.