Cumulative lineup height RAPM

[J.E.]

Ran RAPM with nothing but cumulative lineup height as variables. The results aren't really surprising. Bigger lineups are better on Defense, smaller lineups are better on Offense

Expected difference, on defense, from smallest to biggest lineup, is ~6.2 per 100 possessions. The effect is less drastic on offense (~3 maybe)

[Statman]

Ran RAPM with nothing but cumulative lineup height as variables. The results aren't really surprising. Bigger lineups are better on Defense, smaller lineups are better on Offense

Expected difference, on defense, from smallest to biggest lineup, is ~6.2 per 100 possessions. The effect is less drastic on offense (~3 maybe)

Pardon my ignorance, I always get the various +/- stuff confused. I just want to be more clear on what I see.

This is prior informed +/- with no box score data whatsoever (this is my assumption)? None prior informed? This is statistical +/- melded with real +/-? Is this the one that even includes player height in the defensive credit when computing the results?

I've always felt statistical +/- (the ones I remember seeing here, and what I assume I'm not looking at in this thread), when divvying defensive credit, often put way too much credit to big men (I think because of D rebounds involved in the D calculations). Is that completely unrelated to what you are doing here, or could that be part of why these results look the way they do?

Again, sorry if I'm just way off base here if there would be no bias whatsoever by the nature of the metric, I'm just wondering how much the height factor may be bias with the metric you are using.

I would expect taller lineups to be better on D, smaller ones better offensively - my expectation is that it would for the most part even out (I've long gone to the conclusion that tradition lineup construction is greatly outdated). Your results suggest otherwise, that "big" lineups, in general, are too valuable defensively to be ignored.

Maybe I shouldn't be surprised - if given two identical players in terms of skills - the one that is taller & longer would be the one any of us would want, mainly for defensive reasons. That should play out over the larger scope of this type of study, that the general overall skill of these lineups across all heights should be about the same, but that height tips the scales. That being said, all of us would still take the curry/Klay/Draymon/Iggy/Barnes lineup over some general big lineup any day, because of how skilled that small lineup is together easily offsets any possible defensive post & rebound issues.

Finally, what is the BIGGEST lineup that had some real playing time together that you came across, if that info is easily found? How about the smallest? I'm just curious what the height extremes of actual NBA lineup construction look like.

[colts18]

Can you post that chart for RAPM by lineup height difference so that we can see the expected offense/defense for a lineup 10 inches smaller than opponent or one 15 inches taller than its opponent.

[Crow]

Cumulative weight might be interesting. Does it vary much from height? Age?

Offensive RAPM by cumulative average usage and by cumulative average assist rate might be interesting too. So many different RAPM variants that could be run.

[J.E.]

Statman, this has nothing to do with statistical +/-. It's Adjusted Plus Minus (regularized), but ignoring who the players are, and instead of using the players as variables, uses cumulative height of the lineup as only variable.

E.g. possession #1 might be: 400 inches v 388 inches: Scored 2 points. Etc.
No other info gets used

Finally, what is the BIGGEST lineup that had some real playing time together that you came across, if that info is easily found? How about the smallest? I'm just curious what the height extremes of actual NBA lineup construction look like.

It's 2005/06 Dallas
Podkolzin (7-5), Diop (7-0), Mbenga(7-0), Marshall (6-9), Josh Powell (6-9). Looks so comical that I think my parser may have mis-parsed who's playing

Can you post that chart for RAPM by lineup height difference so that we can see the expected offense/defense for a lineup 10 inches smaller than opponent or one 15 inches taller than its opponent.

I'm a little confused, because I already posted a chart, and I think you can derive the numbers you want from that. Maybe I'm misunderstanding what you want

[v-zero]

This done using the geometric mean of heights rather than the cumulative heights would be really interesting, not that it isn't interesting as is.

[xkonk]

JE, how does this fit in with your post on small-ball vs twin towers lineups? My memory is that the effects there were more symmetric, whereas here there would be a net benefit for playing tall lineups (i.e., the spread you mention here of ~6 points for defense vs. ~3 for offense).

[J.E.]

JE, how does this fit in with your post on small-ball vs twin towers lineups? My memory is that the effects there were more symmetric, whereas here there would be a net benefit for playing tall lineups (i.e., the spread you mention here of ~6 points for defense vs. ~3 for offense).

I think it's apples and oranges because we're looking at 2 different things, one of which has rather arbitrary cutoffs: In the small-ball analysis I used 6-10 as a cut.
A lineup could have been labeled as small ball even when it consisted of 5 players that are 6-9

[kggk]

This is only slightly related, but it reminds me that I've wondered if teams should be far more aggressive substituting for offense vs defense. Every time there is a dead ball, a team could substitute based on whether they are starting the next possession on O or D. If there happens to be an odd number of possessions before the next dead ball, they could benefit by having their more specialized O/D players on the court for an extra O/D possession. For example, if the Pelicans did this all game, it could add up to having Asik out there for several extra D possessions, and Gordon for O.

Of course, this would result in substitutions every dead ball, and the players would hate it, but would it work in principle? How many substitution opportunities don't currently get used?

[J.E.]

This is only slightly related, but it reminds me that I've wondered if teams should be far more aggressive substituting for offense vs defense. Every time there is a dead ball, a team could substitute based on whether they are starting the next possession on O or D. If there happens to be an odd number of possessions before the next dead ball, they could benefit by having their more specialized O/D players on the court for an extra O/D possession. For example, if the Pelicans did this all game, it could add up to having Asik out there for several extra D possessions, and Gordon for O.

Of course, this would result in substitutions every dead ball, and the players would hate it, but would it work in principle? How many substitution opportunities don't currently get used?

Well, other research I've done (posted here somewhere) suggest that players play better when they play X consecutive minutes. I think playing 6-8 minute stints was best. That would probably offset what you suggest

That said, I'm often baffled why coaches don't do it more at the very end of the game (last 2 minutes), when there's constant stoppage anyway, and no potential to even play long stints

[mtamada]

[...]

Of course, this would result in substitutions every dead ball, and the players would hate it, but would it work in principle? How many substitution opportunities don't currently get used?

Well, other research I've done (posted here somewhere) suggest that players play better when they play X consecutive minutes. I think playing 6-8 minute stints was best. That would probably offset what you suggest

That said, I'm often baffled why coaches don't do it more at the very end of the game (last 2 minutes), when there's constant stoppage anyway, and no potential to even play long stints

We do see exactly this sort of subbing at the end of games, when coaches with a lead will send in players with better FT shooting and ballhandling skills, knowing that they will either get fouled or will want to play keep away. Teams that will be on defense after a timeout with seconds left in the game will put in their defensive specialists, knowing that the other team is going to do nothing but go for the last shot. Or the team on defense will put in a scrub with few fouls to go out and deliberately foul a player.

Of course this only happens at the very end of close games, when the gains from using those specialists are clear. Why not do it more often, e.g. in the last 2 minutes instead of the last 24 seconds? We do sometimes see such specialized substitutions, I'd have to guess that the reason we don't see more of them is the reason you already alluded to: with the game lasting more than mere seconds, the coaches want to get the advantage of having warmed up players on the court, not players who've just jumped off the bench to get into the game.

[mystic]

It's 2005/06 Dallas
Podkolzin (7-5), Diop (7-0), Mbenga(7-0), Marshall (6-9), Josh Powell (6-9). Looks so comical that I think my parser may have mis-parsed who's playing

Your parser works fine. As it seems they played that lineup in a game against the Clippers for the whole 4th quarter. Here you go: http://www.basketball-reference.com/box ... 90DAL.html

[J.E.]

Thanks, mystic

Ran a similar analysis, but this time using individual height instead of cumulative height

[tarrazu]

I did see the RAPM player pace effect awhile back from JE, but could you show how much faster or slower a team might play, using cumulative height as the only variable?