APBRmetrics

Focus on best lineups vs. dink lineups

Warriors top 5 performing lineups account for 67% of team points edge. Top 10, 91%. Top 20, all of it plus 102 more points. That is, the bottom 470 lineups on the season's best team combined are -102. Overall the best or second best team in regular season of all-time is making positive results on only 33% of its lineups. Boo dink lineups. Yay concentrating on the best.

For lineups with over 200 minutes played:

http://www.basketball-reference.com/pla ... y=diff_pts

GSW has #3 and 4
OKC has #2 and 6

Toronto has #1 and #13
Cleveland has #7

Clippers have #9 and 11
San Antonio has #12


Houston's best at 27 is not near good enough. Boston at #29, same thing. Atlanta at #23, meh.


For the future, it is worth noting Utah has #5 and 16, while Charlotte has #10.

Crow wrote: Warriors top 5 performing lineups account for 67% of team points edge. Top 10, 91%. ...

Are these the 5 and 10 most used lineups, i.e., minutes together?
If you're looking at the best lineups by point differential... <-- this isn't known until after they've been used.

There are different ways to look at things but the GSW lineups were "best" as stated. By best though I should clarify that I mean highest net points for season, not best per 100 possessions. Still the best 5 were all top 6 in minutes used. 7 of the best 10 were top 10 in minutes used. 13 of the best 20 were top 20 in minutes used. The top 20 represented 48% of all minutes played.

They became "known" after the first 10-20 games and more and more every day the results for the season to date are updated and available for consideration of what to do from there.There is some change but I think most of the best have probably been that way for awhile. Yes the data changes but you can either keep up with it, play the best and adjust as needed or you are dink away a large share of total minutes of lineups with far, far data, and far, far less certainty.

The best 20 lineups averaged +26 points per 48 minutes. The other 470, -2.6 per 48. The data is pretty darn clear and imo sufficient to say play the best (as they are known at that time) as much as possible and the rest as little as possible. And adjust daily or weekly if /when the data suggests you should.

The Warriors are great but probably still sub-optimal, given the use of 470 mostly non-positive dink lineups and plenty of options within feasible rotation limits to use them less or perhaps better.

Celtics vs. Warriors recently: Stevens' starters (with modest productivity traditional starters not available due to Crowder injury) plus 9. Everything else he ran out there, a net -6. Fortunately he played a starting lineup far better on average than he would have if given the chance to play his traditional preference and he played them a good amount. Just barely enough.

Stevens goes back to the meh traditional starting lineup, as expected.

Game 1, Stevens' traditional starting lineup goes -7 in first and only stint and Celtics play uphill the whole game, barely losing.

Game 1, 27 lineups in 48 minutes. Like last playoffs. Less than 30% got positive results. But punch those buttons.

5 playoff games so far, never won an opening stint and lost 2 badly. 5 losses.

Another bad opening stint loss and playoff loss, Stevens' worst yet.

Coaches active this season who started out 0-6 or worse in playoff career? There aren't any, though Steve Clifford could tie that tomorrow and a number never made playoffs to test it.

Injuries are a big factor but after using 27 lineups in game 1 you come back with repeat use of just 4 and then 20 new ones?

It's almost as if he's trying to new lineups because nothing has worked so far.

Game 1 was the closest he has gotten to a playoff victory.

Of the 4 best lineups out of the 8 most used in regular season, Stevens used 2 for more than 2 playoff minutes. 5 minutes of his biggest minute great lineup AB JJ KO MS ET. Unfortunately it was negative, but, come on, just 5 minutes? AB JC JS IT ET got 9 minutes and was positive.

In summary so far it is meh overall traditional starting lineup that usually sucked against good teams -7 nominal (-22 pts per 100 possessions), 2 most used of their best 4 big minute regular season lineups break even and everything else -11 in almost 2/3rds of the minutes. Almost everything about this distribution is not good. Their best got 15-20% of the minutes, the meh starting lineup got close to that in one game and he dinked up the rest of the time to poor results.

It will be hard to win without Bradley, in all of the 4 best big minute lineups.

There are about 4000 minutes to use in the regular season. Finding a positive non-Bradley lineup seems like a reasonable task within that time, among others. Stevens' only found one such positive lineup in at least 50 minutes of testing. JC AJ MS JS IT. Was one considered adequate? Was it not even a formal goal? How much did he use that so far, in this time of great need and turbulence? 3 minutes. Had to try many other non-Bradley lineups that either were less tested, less productive or both. It was negative but just 3 minutes? Elaborate button-pushing may look like genius when it works but when it doesn't it looks like illogical, grossly sub-optimal based on the data you have chaos from a macro perspective. But I haven't looked matchup by matchup for situational logic. That would take more time. Perhaps that would look more plausible. But it still doesn't look good on the scoreboard. Yet.

Was the traditional starting lineup good against the Hawks, making it appropriate despite the meh overall performance and the poor performance against good teams? No. The starters were horrible against the Hawks in regular season, -29 pts per 100 possessions. But somehow Stevens decides to roll it out there anyways and cost himself game one.

Playoff lineups used 10 minutes or more http://bkref.com/tiny/nSCkY
Raptors with #2 and 3. Spurs get 5th. Warriors only 18th. Indiana with a top 10 and 3 of the worst. Celtics get 25th of 33. Mavs, #1 and 3 in or almost in bottom third. Thunder with #8 and 11. Cavs #9. Clips 4, Heat 7.

Stevens goes to a dink lineup (used just 17 minutes for season and performed very well) to start game 3 and gets a plus 9 opening stint and holds on to a plus 4 for the game, contributing to the win. Uses it again in game 4 and it goes -12 but wins off a big plus stint from another lineup. Was this starting lineup a good choice? It worked the first time and could be called a fine coaching move, though if an outsider suggested going to a lineup that was great in 17 minutes many would call that a real sketchy or even silly recommendation. Can bend it either way you want. Better tested would be more reliable but we'll see what happens next. Winning off one or two great stints and different ones game to game is different than pounding out edge consistently with same lineups but a win is a win.