Maybe the only thing the defensive positional averages are saying is that when bigger positions are subbed in for smaller positions, the defense improves. And the offensive positional averages are saying that when teams play small, they are better offensively.

Generally adjusted plus/minus has a lot of issues with players moving between positions.

How much are positional averages affected by the accuracy of positional assignment and how much by the use of non-traditional lineups (2 PGs together or no designated C or 2 SFs, etc.)? The two are related and there probably isn’t a definitive answer agreeable to all. But I wonder how well unconventional lineups perform on average, adjusted.

Looking at 05 Rosenbaum defensive study and this list, a few curious cases:

Mark Blount went from one of worse starting defensive centers to one of best? Change in rank related more to change in team and team context or his dramatically better defensive play?? “Defensive rating” went the other direction (heavily because of the different teams involved)

Dampier went from one of best to one of worst? (Change in coach, pace, subs)

Jamison from one of the worst to one of the best.(Change in coach, pace)

Battier from 1st to just average? (change in team but not pace or team defensive quality and even the same sub much of the time Wells). I can accept some slippage but error may be overstating the change.

Banks from 2nd best to 11th worst.

Probably more cases of agreement then disagreement (I saw a fair number in agreement in my spot check) and there can be several reasons for disagreement of course. But worth a more thorough check.

Your intended study of team change cases could help. Change of coach an dpace may be other useful studies. Change of teammates and especially sub of course has impact too.

My impression is this general +/- approach does a better job of finding the top and bottom players than sorting the middle or finding the magnitude of impact of the best and worst.

Dan Rosenbaum’s net statistical +/- can be calculated and has been broken into offensive and defensive sides. You’ve indicated a desire to calculate your own. Using either and your newly available “pure” offensive and defensive adjusted +/- it will then be possible to split the offensive and defensive impacts into 4 boxes - those produced as the lead actor and those “produced” as a member of a team / 5 man lineup by subtracting the net individual statistical +/- part from the pure total for offense and defense.

I think the same approach developed here to find offensive / defensive impacts could be used to address each of the 4 factors, constructing something similar to DiffOD for FGs made, turnovers, FT made or frequency and rebounds on offensive and defensive side of action. What do you think?

If this were done and net statistical +/- were broken down into 8 parts as well then subtracting the parts of statistical from the pure parts you’d have a 16 aspect portrait of a player’s individual and team impacts. Something I’ve pointed to as a goal now within reach, to the public.

Look forward to the further advances you are leading.

It is pretty close but the edge goes to the front court having more positive impact than backcourt.

Instead of it being a little man’s / PG league right now, now more than ever according to some based largely on the offensive performance / press of the top PG names, PGs as a group are having the least positive net impact of any position across the league, by this measure. 2nd best on offense but last on defense.

Somewhat surprisingly that SF came out best.