Adjusted 5 man units

[J.E.]

Are you bored with adjusted player combinations yet?

Lambda = 500. Lineups with less than 180 possessions total not listed. 180 is approximately equivalent to one full game
http://stats-for-the-nba.appspot.com/5mu_11 with playoffs and
http://stats-for-the-nba.appspot.com/5mu_11np without.

CTRL+mousewheel is your friend (to get all lineups fitted in one line)

Sample size isn't great, so take everything with a grain of salt

Obviously the top is dominated by Boston, LA, Dallas, SA, Miami

Chicago's best unit was Deng, Brewer, Rose Gibson Asik +13
Chicago's starting unit Noah Rose Bogans Deng Boozer with a +5.3

Best OKC unit was Collison Durant Westbrook Harden Ibaka with +7.6
Their starting unit with a 0.4

Boston's "big 4" don't care who they're playing with. They look good with Shaq, Jermaine, Davis, Krstic and Perkins

Miami with no "super" unit but several really good ones.
One unit with the big 3+Bibby+Ilgauskas didn't play much in regular season but did in the playoffs and got completely destroyed (-7.1)

LAL with Artest Odom Gasol Fisher Bryant at #4, substitute Barnes for Artest and they became horrible (in limited minutes)

LAC with Davis Griffin Gomes Jordan Foye at +6.1. Drops to -5.4 with Williams instead of Davis. Probably not the best trade the Clippers ever did, even moreso when you consider that Cleveland gets Kyle Irving too

Surprises are
Utah at 5 with Kirilenko Williams Jefferson Miles Millsap in limited minutes. Was blown up
PHI at 10 with Iguodala Holiday Young Brand Meeks
HOU at 12 with Martin Scola Battier Lowry Hayes. Also blown up
IND at 13 with McRoberts Hibbert Collison Dunleavy Granger

[Crow]

Nowitzki+ Chandler + Stevenson + Kidd + Butler. which had the best RAPM here, also had the best traditional APM estimate for lineups over 150 minutes.

I'm going to look at some other lineup rating comparisons. The scale of RAPM seems much more credible.

[DSMok1]

Excellent!

I actually have already run these myself, but you'll save me the work of posting them! I agree with your selection of lambda; that's what I came up with as well.

I guess I'll have to move on to using informed priors to do something original!

[J.E.]

Multiyear analysis online

http://stats-for-the-nba.appspot.com/5mu_allnp

Best unit was -Kidd Terry Howard Bass Nowitzki-
followed by -Rondo Allen Pierce Garnett Perkins-

Two Orlando units made it into the top 6. They weren't happy with either one of them and made changes to both in the following offseason

Lakers had surprisingly good units in
-Farmar Bryant Vujacic Odom Gasol- and
-Fisher Bryant Walton Odom Gasol-. The latter with a whole bunch of minutes, too

The Warriors had a really good unit with many minutes
-Davis Ellis Jackson Harrington Biedrins-, and another good one with Pietrus instead of Harrington in less minutes though

-Paul Peterson Stojakovic West Chandler- was a good unit

Toronto had two good units
-Nesterovic Bosh Moon Parker- and either Ford or Calderon

Bad units:
-Rondo Allen Pierce Davis Perkins-. Tells you something about KG

worst unit overall was
-Flynn Brewer Gomes Jefferson Hollins-

-Ellis Curry Wright Lee Biedrins- was not a good unit but was used a lot nonetheless. Subsitute Udoh for Biedrins and it's a good unit

-Westbrook Sefalosha Durant Green Krstic- with an enormous amount of minutes but just an average rating

[EvanZ]

Jerry, I was looking at 4-yr RAPM (for individuals) and noticed that Durant is only 1.1. That really doesn't seem right considering his single year RAPM numbers at least the last couple of year. What do you think is going on?

Ah, nevermind, he had a couple of really bad seasons to start his career.

[Crow]

Thanks for another desired RAPM product.

"Big" minute, high achievement multi-season lineups are pretty rare.

Only one third of this top 436 performing lineups exceeded 1000 possessions for this extended timeperiod.

Only 10% exceeded 2000 possessions. About 85% of the teams with these lineups won 50+ at least at some point during the period.

Just 10 lineups exceed 4000 possessions. All but one of those won 50+ games at least at some point and Memphis might next season.

A top performing 5 man lineup is a pretty big part of the quest to be a top team but it is not always a clear-cut, easy thing to assemble and maintain even with some good and great players and some helpful role players. They generally have to at least fit pretty well and hopefully they enhance each other.

[J.E.]

Durant

His single year RAPM rating makes the biggest jump ever seen from 08/09 to 09/10. His BoxScore numbers make the jump one year earlier.
In 07/08 and 08/09 he was minute leader on teams that won 20 and 24 games. That usually doesn't sit well with RAPM unless you have a good On/Off. In 07/08 they were -11 with him on, -2 with him off. In 08/09 they were -8 with him on, 0 with him off ....

[Crow]

Lineups +3 or or better overall (offense and defense) for conference finalists

Bulls 11 for 6395 possessions
Heat 7 for 5253 possessions
Mavs 7 for 4285 possessions
Thunder 5 for 3054 possessions

All those lineups and just these few estimated strongest ones.



For lineups above +5

Heat had 5 that benefited from (I think) an average of 1.0 offensive trios (Wade-James-Bosh in every case) over +1.5 and 2.4 defensive pairs better than +1.5.

Mavs had 5 that benefited from (I think) an average of 1.4 offensive trios over +1.5 and 2 defensive pairs better than +1.5.

The best lineups have a decent amount of best offensive trios and best defensive pairs as building blocks. I didn't calculated the average sum of the contributions of these trios and pairs to the average lineup strength but it is probably pretty high. It appears that the data sets are fairly consistent and may be showing something of value.