Strong lineups

[Crow]

The top candidates for strong or "super" lineups based mainly on raw +/-performance to date (but also looking at APM) in at least 50 minutes of play belong to Portland (maybe, not as impressive as the others), Indiana, Miami, the complete OKC bench unit, Lakers and Chicago.

http://basketballvalue.com/topunits.php?year=2011-2012.

All these lineups have 4-5 guys returning from last season. And yet for 3 of the 6 lineups, the guys who were around last season did not play much and / or well together then (Portland, Indiana and Lakers). Either the lineups were not discovered last season, or they weren't strong last season but are now, or might not be as good this season as they currently look. Will want to check them again mid-season, after the full season and after the playoffs. Dallas was an example of a team that really benefited greatly from a super lineup or at least a super performing lineup in the playoffs last time.

[DSMok1]

J.E., when will you update your 5-player lineups for 2012? The 2011 w/ playoffs version is here: http://stats-for-the-nba.appspot.com/5mu_11 .

Interestingly, most of the best lineups from last year have been broken up. Chicago's top lineup is still intact, as is Miami's. Other than that, I don't think any of the top 10 are still on the same team together.

[Crow]

Philly can still field 10th best performer at that link page. Boston 11th best. San Antonio and the Lakers still have lineups that were top 20. OKC one just outside the top 20.

[J.E.]

J.E., when will you update your 5-player lineups for 2012?

I stopped updating these once I found out that simply adding individual player RAPM values gives better prediction results than adjusted 5 man units. For this analysis to be in any way useful I will probably need to create a prior for the unit using individual RAPM values, then penalize for moving away from that. I've done no research on that front though, so I don't know what lambda is gonna be like (might be infinite), so, for now, I encourage everyone to simply add player RAPM when estimating production of a lineup

[DSMok1]

J.E., when will you update your 5-player lineups for 2012?

I stopped updating these once I found out that simply adding individual player RAPM values gives better prediction results than adjusted 5 man units. For this analysis to be in any way useful I will probably need to create a prior for the unit using individual RAPM values, then penalize for moving away from that. I've done no research on that front though, so I don't know what lambda is gonna be like (might be infinite), so, for now, I encourage everyone to simply add player RAPM when estimating production of a lineup

I'd love to see you do that. Could be very, very useful. Would show synergies and benefits of good lineup composition.

[Crow]

Simply adding individual player RAPM values did a good predicting results for big minute Dallas and Miami lineups I checked from last season.

But this method didn't predict anything close to what the best Utah lineup accomplished together last season and it didn't predict what the best Indiana lineup did last season and it doesn't appear to be doing to their best big minute lineup that still has Collison / Granger / Hibbert lineup this season either. Certain Memphis and Clipper lineups far exceeded sum or player RAPM expectations too. The exceptions are worth looking at more closely to try a rationale for the discrepancy. Some performances may be flukes and you would expect some even despite the fairly large minutes, but some may be real and based on the strong chemistry / completeness of that particular lineup well above others.



For what it is worth I share DSMok1's interest in your proposed new lineup method.