APBRmetrics

[Mountain]

82games.com provides this file
http://www.82games.com/USORT13.HTM

Sorting by minutes for season I see only 6 5 man lineups in the entire league were used for an average of more than 10 minutes a game. 5 of those 6 won an average of 57 games and the Sixers surprised.

No lineup was used over 15 minutes a game.

Only 18 lineups were used for over 5 minutes on average.
Given a couple of teams with 2 such lineups about half the league had no lineup they used on average 5 minutes a game.

Teams averaged only 3 lineups that they used more than 2 minutes a game, 164 minutes total.

Injuries, matchups, trades play a role of course as I noted when I addressed this topic in the past but this is still bizarre in my mind so I raise it one more time.

Outside the 18 lineups played over 5 minutes a game or maybe a little deeper I don't think you can really say much with any confidence about the effectiveness of the sea of other lineups.

Only 2 teams played 2 lineups for more than 5 minutes a game and they varied by only one player. There is no statistically meaningful comparison happening. No coach, GM or consultant has a decent database from NBA game play within a season for choosing which 5 man units to put on the floor.

If the NBA believes 5 man units are the fundamental unit of study the coaches sure don't make it easy to study and adjust. 69% of all team minutes go to lineups used less than 100 minutes for a season. Most teams split their time into 150-250 tiny samples. Why do they manage minutes in a way such that they leave themselves pretty blind to effectiveness at this level? If you believe 5 man units matter you could study 5-10 lineups a good deal better than they are right now.

Given the management of rotations it appears to me that coaches believe more in systems and coaching than they do in the synergy of 5 man lineups. As important as systems and coaching are I think this is a significant self-inflicted handicap.

But after saying all this the answer may be that they feel they can make these choices well even without a statistically significant database and they have some data in their favor.

The lineups the league coaches give 100+ season minutes to have an average per 48 minute point differential of +3.8. The lineups given over 2 minutes a game win by the equivalent of 4.7. Lineups given 5+ minutes a game win by 5.8. The cream of crop lineups given 10+ minutes a game win by 7.8.

But this doesn't get to the missed good 5 man lineups not recognized and promoted. Maybe a few coaches catch a few good ones but without a better database it is a tough judgment or guess. This good data should strengthen the case for disciplined search for good performing 5 man lineups.

[Mountain]

Eli has posted a major new study provided adjusted lineup rankings.

http://www.countthebasket.com/blog/2008/04/24/adjusted-lineup-rankings/#comment-196

It deserves its own thread but his study allows the above topic to be addressed using possessions instead of minutes.

I see the league used 356 5 man lineups at least 100 possessions or on average about 12 per team.

Altogether the league used 10,871 lineups.or on average 362 per team.

The lineups used at least 100 possessions got 42.7% of the possessions. 57.3% went to lineups used under 100 possessions.

Almost 25% went to lineups used under 20 possessions (less than a typical quarter) over the entire season. There were 8700 such lineups. On average team coaches ended up using 290 different lineups for such brief intervals.

Only 31 lineups in the entire league got used 500 possessions or about 1 per team.

It surprises me that so few lineups are well-tested, trusted and sustained.

One approach largely treats the game as the sum of individual contributions (with the exception of defense). Perhaps the data lends itself to that theory because teams are not managing lineups in a way to maximize the significance of the evidence about the efficacy of 5 man lineups and may not getting the best possible returns.
Perhaps player synergy is not showing up very strong because it is not that finetuned. Seems like a wide open frontier for research and more disciplined lineup management.





Multiply totDiff by minutes played and the 3 specific 5 man lineups providing the most "edge" in the league are by far on Boston, Detroit and New Orleans.

The lineup that contributed the most cumulatively to negative point differential over its minutes? New York's infamous group with Marbury and the Curry / Randolph pairing.

.