2018-19 lineup analysis

Home for all your discussion of basketball statistical analysis.
jimmyp
Posts: 1
Joined: Tue Mar 19, 2019 6:57 pm

Re: 2018-19 lineup analysis

Post by jimmyp » Thu Mar 28, 2019 7:34 pm

How much of the lineup effectiveness is due to the individual players' effectiveness and how much is due to synergy, for lack of a better word? Assuming that it's not a ridiculous lineup, it seems that the first would dominate the second. I'd like to read more about it but can't seem to find any published research or posted data.

Crow
Posts: 6061
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2018-19 lineup analysis

Post by Crow » Thu Mar 28, 2019 10:20 pm

Looking at player RAPM / RPM and 5 man lineup RAPM might give some clues but maybe not reliably. Add sublineup RAPM. Try to sort it out.

Or ideally run RAPM on players and pairs in the same process.

Crow
Posts: 6061
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2018-19 lineup analysis

Post by Crow » Mon Apr 08, 2019 7:20 pm

Only 4th time in 19 seasons that no lineup will get 1000 plus minutes. Imo there should be at least 10-15.

Less than 1 per team over 300 minutes. Down a bit from last season.

Almost 20,000 lineups used across the league. Up about 13% from last season.

Lineups getting less concentrated by these measures. Absurd.

Percent of lineups cumulatively positive fell from 32% last season to under 30% this season.

Only about 5.5 lineups per team got 80 plus minutes and were therefore "non-dink". 61% of these were cumulatively positive. For dink lineups under 1 minute per game for season only 25% were positive.

If you are a GM or owner and interested in this topic, message me. Your Coach and analytic staff are probably not paying enough attention to these trends. You might be doing well but it is based on talent differential, other stuff or somewhat less inefficient lineup management compared to the competition. But you probably could do better. On average a non-dink lineup is about 7 points better per 100 possessions than the average dink lineup. If you are not interested in change and trying to capture some or all of those apparent 7 extra pts / 100 poss. then carry on with current coaching practices... It seems like hardly anyone is aware than more 60% of total time is given to much worse performing dink lineups. I've tried to call attention of this massive inefficiency but to apparently little avail.

Crow
Posts: 6061
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2018-19 lineup analysis

Post by Crow » Tue Apr 09, 2019 12:32 am

Kings, 612 lineups used. 33.5% were cumulatively positive. Better than average but not that impressive. 4 most used (over 1.5 minutes per game), all around neutral. Not really any bad big minute pairs but not really any strong ones either. The best ones of the lot will disappear if WCS leaves.

Crow
Posts: 6061
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2018-19 lineup analysis

Post by Crow » Tue Apr 09, 2019 12:38 am

Lloyd Pierce "has done a good job"?

Of 7 most used lineups, 4 have performed horribly, one just moderately bad, 2 around neutral. Only 4 lineups over 1 minute per game and none over 4 minutes / gm for season. 19 of 20 most used player pairs negative. 7 biggest Young pairs all from moderately bad to quite bad.

988 lineups used. May hit 1000. Only 26% positive.

Horrible defense. Offense has moved from horrible to pretty good. I am not sure they stay there or get better.

Crow
Posts: 6061
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2018-19 lineup analysis

Post by Crow » Wed Apr 10, 2019 5:05 am

I'd ask a coach to design a 8 lineup rotation for the first half. For the playoffs (and for regular season). You can adjust the minutes and time of insertion based on matchups and performance.

For the rest of the conceptual plan... be a team and contact me.

Crow
Posts: 6061
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2018-19 lineup analysis

Post by Crow » Thu Apr 11, 2019 11:02 pm

JB Bickerstaff used 973 lineups this season. Only 30% positive. Had a plus 7 pts / 100 poss. starting lineup. Could have used it more. Mistakes on both ends. Only 1 of the 973 lineups used over 1.5 minutes per game for season. 1. Some promising lineups not tested enough because of the incessant change from one dink lineup to the next, next, next ad nauseam.

I dunno if the fired execs went along with this chaos, to tank or not willing to direct for more concentration; but it didn't happen.

Are you going to be different as GM Zach Kleiman? Let the next coach do pretty similar or demand very different? If you want different, I could provide some feedback / help. You've seen what the dink parade produced.


I really don't think many people have recognized what coaches are doing. To get to 973 lineups used, Bickerstaff had to use about a dozen new, never used before lineups EVERY SINGLE GAME OF THE SEASON and then used maybe a dozen or more that he has used before but most only a few times. If you know the team and what you are doing why do you need to go with 973 different approaches??

Crow
Posts: 6061
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2018-19 lineup analysis

Post by Crow » Thu Apr 11, 2019 11:59 pm

Brad Stevens "only" used 522 lineups. Far less than when I first checked him a couple years ago. 37% positive. That is very good, compared to the competition.

But... only one lineup used over 2 minutes per game for season and only 4 more over 1 minute.

Still plenty of room for more concentration on better, bigger minute lineups and less spraying of dink lineups that are non-positive in almost two-thirds of the cases. The better can get still better.

As it stands, the Celtics go into playoffs with a pretty good idea what the starting lineup has done. But, oh wait, Smart is out. So there are left, at least for a few weeks, with very thin guesses what the next 4 lineups tested 0.75 to 1.5 minutes per game for season might do (in a new context) and thinner and thinner guesses about everything else. Well that season wasn't so useful at gaining reliable knowledge compared to what they might gotten with more disciplined lineup management and research strategy. Oh well, wing it. Stevens is better at winging it than most but it is still winging it for high stakes.

DSMok1
Posts: 899
Joined: Thu Apr 14, 2011 11:18 pm
Location: Maine
Contact:

Re: 2018-19 lineup analysis

Post by DSMok1 » Fri Apr 12, 2019 1:16 pm

Here is the final lineup regressed net ratings for the 100 most played lineups from @bbstats (Nathan Walker):

Image
Developer of Box Plus/Minus
APBRmetrics Forum Administrator
GodismyJudgeOK.com/DStats/
Twitter.com/DSMok1

Crow
Posts: 6061
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2018-19 lineup analysis

Post by Crow » Fri Apr 12, 2019 2:59 pm

Thanks. I'll look at it. Getting beyond raw plus minus is obviously important when it can be found or produced.

Of the top 20, 4 for Toronto, Portland and Golden State 3, Philly 2. That is 60% of the spots. More than one is good flexibility. No one else over 1.

Warriors best lineup barely used more than 2 minutes per game. The regular season is just a hassle to get thru. In last playoffs it was used more than 3 times as much. Still low. Didnt need more apparently. Only used it in 2 games in finals. But then at almost 10 minutes per game. Hammer down, opponent goes down. Concentration sparingly applied. Just too easy from them. Nobody else had the luxury to under-utilize their best. But most probably did.

Crow
Posts: 6061
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2018-19 lineup analysis

Post by Crow » Sat Apr 13, 2019 3:19 pm

Rockets, overcame slow season start... but great surge only nets them the toughest home court slot.

At lineup level the 6 most used were 1 great, 2 good, 2 meh, 1 terrible. As is apparently the norm now, Rockets only had 2 lineups over 1.5 minutes per game for season. The starting lineup was good but only used 5 minutes per game for season. If the starting lineup becomes unavailable, hardly anything has adequate testing to replace it. That is true in general too, of course.

653 lineups used, just short of average level. Just short of 35% cumulatively positive for season. That is notably above average, though it still sounds not that hot and pretty random to me. About 70% of total time went to dink lineups. For all the analytic horsepower and activity there, it is very traditional, not particularly impressive lineup management. Could use more focus & push there.

Crow
Posts: 6061
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2018-19 lineup analysis

Post by Crow » Sat Apr 13, 2019 3:37 pm

Which playoff team lacks a top 20 performing bigger minute lineup on this list? Celtics. Bucks. Uh oh. And 4 others. Clippers and 3 lower seed eastern teams.

Crow
Posts: 6061
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2018-19 lineup analysis

Post by Crow » Sat Apr 13, 2019 3:52 pm

How did Bucks not have a top 20 performing bigger minute lineup here? They have a big minute starting lineup that was plus 7.4 pts / 100 poss., but I guess the strength of the opposing lineups must have been weak and pulled down the adjusted performance. Rut roh. Not good to start playoffs without a proven big minute lineup that is a real hammer.And injuries too.

Plan B? No champagne. Only other lineup that barely got 1.5 minutes per game for season was horrible. NOTHING ELSE used over 1 minute per game for season.

Apparently Budenholzer loves to play the dink game. 816 lineups used, or about 20% more than average. 35% positive is good but not that great for top team. An impressive 85% of his 20 most used lineups were positive but they only got 35% of total time. That is not enough concentration imo. Not anywhere near enough.

Is it harder to win with the dink game in the playoffs? I'd assume so. Biggest minute lineups generally play more and decide more of the outcome.

Crow
Posts: 6061
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2018-19 lineup analysis

Post by Crow » Mon Apr 15, 2019 8:13 pm

For Thunder - Blazers game I tried to do some more detailed lineup matchup analysis. Unfortunately the BRef and Gameflows data wasn't in perfect synch. It is also a bit hard to do this manually. But here are the quick findings:

Both teams played an average of 3.5 to 3.6 starters over the course of the game.
For over 80% of the time the Thunder faced 3 plus starters.
When both teams had 3 plus starters on the court the Blazers won decisively. By at least 12 points.
When Thunder had 3 plus and faced less than 3 starter Blazers, they won modestly.
When Thunder had 2 of less starters, they won strongly.

Crow
Posts: 6061
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2018-19 lineup analysis

Post by Crow » Tue Apr 16, 2019 12:11 am

Ideally I would recommend teams try to have a lineup distribution where the starters get 1000-1200 minutes over the season, two variations of the starting lineup which get 400-600 minutes each (heightened research in case you had to go with one of these in lieu of the starters and as the first main variations form the starters), and 3-5 more main rotation lineups that get 200-300 minutes of testing. These 6-8 lineups would get around 75% of all minutes. As much as possible of the remaining minutes would go to 20-50 other lineups that had an explicit rationale as a good response to specific foreseeable situations. Some would go to other dink lineups but only when the team design & performance wasn't working. If you really know lineup construction some will look better than others and you should be disciplined enough to go with your prior expert judgments over spur of the moment reactions and randomness.

Post Reply