APBRmetrics

Pelicans with the best performing lineup per 100 possessions used over 200 minutes.


Redick is not in it. 7 of 10 of his most used lineups are negative. Some badly so. His LA-RAPM is only around -1. A role player til his contract finishes next season. After that? Shouldn't be big money.

For Griz, most player pairs are not far from neutral. The clear best is Clarke-Melton at +12/100p.

Some problems with Jaren Jackson and Kyle Anderson with some starters. Both good with Melton. Jackson isn't positive with anyone else in main rotation. Anderson only with Melton and Clarke.

Melton Anderson Clarke Jackson is great and doesn't need Morant to be great. +35 with T Jones, though in only 48 minutes. Maybe should have been 480 plus minutes test.

5 of ten most used Morant lineups are bad to horrible. 3 of the 5 good ones left with Crowder. The remaining two include Brooks Anderson Jackson or Melton and JVal.

Morant best with Melton. Positive with few others. Near neutral with most. Bad with Allen. Leader? Not really that much, yet.

No remaining lineup used over 2 minutes per game for season. Good and bad performing candidates, none with enough testing. As good as the new coach's general reviews are (and new management), this is a huge strategic mistake. Rotation needs extensive optimization. Not bad overall but appears it could potentially be substantially better. If Griz want a 4-7 lineup rotation for 80 plus % of minutes, ask me for it.

4 lineups (all big winners in limited tests) with 8 players looks like a great playoff rotation. Player minutes work. Quick n easy.

Might not work in every game; but that is where I'd start, adjusting only after adequate trial. Prior tests were only short minutes but I'd go with the best in short tests over lesser performers and / or less minutes. Adding other lineups for more rest in regular season or special situations wouldn't be that hard.

Jenkins picked 40% winners among the 433 lineups used, which was probably about average for this season to its degree of completion. Not bad relatively speaking to others, but not a justification for throwing 433 lineups out there instead of 250, 150 or 50.

Tell me you can't get by with just 50 lineups and I'd say: Try, then tell me what you couldn't do and why that was important. After the 50 most used, everything is 10 minutes or less for season. More than half of these were used less than 3 minutes and none of those more than twice. If they we're important, why weren't they important enough to more than barely use? Count me as skeptical. But I'd allow some exceptions if that bought a meaningful increase in general concentration.

In first 20 games Jenkins tried 183 lineups. Only about 1/3rd of his early guesses were positive early. Over time more got positive. And the rate of new guesses slowed, though he still averaged about 5 never before tried lineups EVERY GAME after the first 20 games of guessing new almost twice as much.

44 of the first 183 lineups got more than 5 minutes in those first 20 games. Around 75% of lineups did not even get 15 seconds per game. The lineups used over 5 early minutes were about 1/3rd more likely to be positive. That is the argument for greater concentration.

Got to take some level of guesses to learn but there should be limits on guesses and something from it that passes- in a stretch- as learning.


There are various distributions of lineup minutes that might be reasonable. An example: one lineup for 800 minutes, 2 for 400 each, 4 for 200 each, 10 for 50 minutes each, 15 more for 20 minutes each and 800 minutes for whatever else the coach wants to try. That would 32 main lineups for 80% of the available time. Spend the rest on a couple dozen or a couple hundred lineups as the coach wishes.

No team gets close to this distribution, which has 7 lineups tested over 200 minutes. Only one team this season has even gotten to 3 lineups tested so far over 200 minutes (Jazz). For the full last season, just 2 did that.

Some say you have to test a lineup at least 200 minutes for the results to be even semi-meaningful. Well, do it 0-2 times like almost everybody or 3 like the current testing leaders... or push up towards 7. That's the choice. And the rational answer is...? Whether to go for 4 or 7, figure it out. More tested adequately vs. some tested even more satisfactorily. But test less than 4 to over 200 minutes? Probably a mistake.

The top of the most used lineups in league isn't as impressive as one might expect or hope. It is 1 great (Jazz), 1 very good (Nuggets) 2 around neutral (Spurs and Nets) and 1 bad (Cavs).

Spurs are a surprise having a top minute lineup. I recall they were usually very low minutes with their most used lineup. Pop picked / managed one that was inadequate.

Nuggets having the 2nd best in these 5 is nice but they don't have as likely an expansion of minutes in playoffs for that lineup. Already at 15 min. / gm. and last playoffs only used that top lineup 15 min. / gm. Could do more and probably should but good chance they don't.

Jazz use their top lineup about 12 minutes per game. Room to expand minutes but will they? They only used the most used about 7.5 minutes per game in last playoffs. Couldn't decide between Ingles and O'Neale in playoffs after a regular season most used split between Favors and Crowder. Some splits might be justified but you have to make it work in the end. Whereas the regular season splits were both very good, the playoff splits were horrible and horrendous against the unusual Rockets matchup. Snyder will have to do better.

Last playoffs 3rd and 4th most used sucked too and then it was into lineups used a total of 10 minutes or less. Probably need more lineup concentration and of course more winners. More disciplined regular season testing might help. Last season that did not extended beyond the 2 most used. This season 3 got tested beyond 2 minutes per game but one was bad. So still dependent on 2. A handful of candidates that look good in under 2 minutes per game but that is too limited to trust. Surprises likely again, bad or possibly good.

The top 2 are a split involving Ingles or Conley with the other starters. The shorter minutes candidates have both, neither or just Conley. With both might be tempting and it has worked in a few lineups but overall that pair is slightly negative and the 2nd worst performing on the 20 most used. Clarkson - Mitchell is the worst (though only mildly negative. Conley - Mitchell is 3rd weakest at short of plus 1. If they lose, the regular season data gives some indication it will be because of failure with guards pairs among the main 4 guys. Going with just one and somebody else might be warranted. I would need to check further to finish this aspect of the analysis. Bogdanovic at SG? Mudiay? Limited data suggests going with Bog at 2 and not Mudiay. Maybe gambling on Bog there a lot, depending on matchup and early results. O'Neale at SG appears to work alright as well.

Just not really 2 guards initiating / creating in the same lineup. Or at least not on average across good and bad lineup completions. Maybe adjust the system or instructions to players to deal with this issue. Not resolved in the season level data. I could look month to month. But I'll either do that later or leave it for them to check their not so helpful dink lineup results.

Mitchell in general is not anything special as a shooter and he sucked at that last playoffs. Suck again and there should be real review about the design and his role in it.

Other contenders have very good or great minutes with less minutes. They have more likelihood of gain from increased minutes. I'd assume most will increase concentration but who does it most / least, best / worst? Who is forced or choses to try some new most used lineup? Lots of questions to be answered with varying degrees of logic and success.

Wizards lineup management:

70% of the most used 20 lineups were negative. Most bad to horrible. No lineup used 3 minutes per game, only 1 over over 2 min. / gm.

80% of 20 most used quads negative. None much over 4 minutes per game. Only positives have Ish Smith and / or Mahinmi. Will any of these be available or used much next season? Were they tested enough? Really thin.

Only moderately strong trio was Beal Bertans and Smith at a still way too light 7 minutes per game. Will that be available or used much next season?

80% of the 20 most used pairs negative and none more than marginally positive.

That is all they could manage or find. A very chaotic, unsuccessful mess.

Smith Beal Bertans and Thomas was ok but only tested about 2.5 minutes per game. Same with these guys and Wagner. Should have been at least triple the time for one or both. Which is better? Is either reliably positive? They don't know. Too busy with less plausible, random dink bull.

Do they use this quad much or at all next season? Surely they will try with Wall instead of Smith but how much? Should be way way more. Will it? Will they try any with Smith AND Wall?

Could really use improvement in lineup management. Need a voice to call out and fight against the unsuccessful, irrational chaos. Or better yet get a rational strategic lineup plan together and implemented instead.

Hachimura's minutes were handled very sub-optimally. Far better player pairs with Bertans and Bonga but only played with them for 40% and 33% of his time respectively.

Not unexpected that he would benefit from offensive stretch from his frontcourt partner and some defensive help on & from the wing.

If they are going to play Rui with a center, they better get a different one and / or use a different strategy.

Napier has far better results with Rui than other guards. Random or doing something more right? Check the video.



Mathews should be retained, given more minutes next season. Maybe more previously if health allowed.

Probably should have given Chiozza more trial. But maybe his amazing team results on floor didn't fit with a tank? Twice as good results as Mathews and light years from the team average. But cut and on to Brooklyn.

8 of most used Troy Brown lineups negative, most horrible. Best performer used less than 1 minute per game. Mostly bad team results as a wing. Some success at PF surprisingly. Should use as a wing way more in a lineup that works. At PF? Maybe, if your other lineup options / choices suck.

Bryant? Matchup data and RPM are neutral but most lineups were bad. None used 3 minutes per game and only 1 over 2 minutes. That is not the way to do it. Test several big and either find / use something acceptable or trade him. Smith Beal Bertans Thomas worked in too short tests with Brown and McRae. Will it work with Wall? Probably, but test the hell out of it early and decide. Would probably mean Rui doesn't start. Or Bonga. Could / should try a handful of lineups a lot and pick & adjust. Could be Bonga or Hachimura over Smith. But check the results.

Beal has worst DRPM estimate of any SG. Better pick it up next year... or collect huge extension money anyways... Hold the team leader accountable for defense? Either didn't work or try harder enough. Help him be more than horrible? Nope.

Heat's 10 most used lineups with Dragic tend to be great or awful. One reason is low usage. Only 1 lineup used over 1 minute per game for season. Best lineups have Robinson and Olynyk but neither is used with even half of his total time. Could benefit from further concentration on best pairs.

Basic lineup philosophy / strategy review:

A team should have a lineup used over 8 minutes per game. It should be one of their best. If it is, it should in most cases be pushed to 12-16 minutes (as long as it continues to work), 20 if possible without other damage.

The lineup rotation should include other lineups tested / used to at least 3-4 minutes per game. If you believe in it, use it, it probably should get at least that much use regularly. Experience together matters, should improve a lineup. 2 minutes is a minimum goal. Less is sub-optimal, a bad practice.

Dink lineups used less than 1 minute per game historically are much weaker performers than those used more or a lot.

The majority or vast majority of minutes should be given to a small number of lineups. 5, 7, 15, 20 (or 50 -75 at the extreme). Using 400, 600, 800, 1000 different lineups in a season is inefficient and unnecessary.

Teams with a lineup even close to 8 minutes per game? 6. Average number of lineups used 3 plus minutes per game? Not even 1.3 per team. About 2.3 over 2 minutes. Ridiculous.

When will this change significantly, for whom and how far? A few are ahead of the pack on biggest minutes lineup or frequency of second tier lineups over 2-3 minutes per game. But very few and they could still go further. 2 made it over 10 minutes per game with most used lineup, something helped by good fortune on injuries. 9 teams have used 3 lineups over 2 minutes per game and 4 more have used 4. The most teams with 3 and over in last 7 years is 15 and the low was 6. The most teams in last 7 years with 3 or more is 5 (twice). That is the frontier for now. I'd like to see 6-7.

Last season the average number of lineups used over 2 minutes was slightly lower but not sure if this is change or affected by this partial season.

The Finals was between a team with 4 vs. 3 over 2 minutes / gm in regular season. That has tended to the case in last few years. Less so the few years before that. In the 2019 Finals itself though the Raptors used 7 over 2 minutes, almost 8. Warriors just 4. Raptors used 33 lineups overall. Warriors 55, affected by injuries but probably still too high. In the Finals the number of lineups used over 2 minutes in the previous 5 years ranged from 3 - 6 for each team with a bit more of the lower end. So Nick Nurse has the recent record for this style of lineup management.

Overall in fairly short minutes, Oladipo was bad with Brogdon, Sabonis and Lamb and good to great with Turner, Warren and the Holidays. Brogdon Oladipo Warren Sabonis Turner is a mashup that works though. Still, Oladipo is Brogdon's worst overall pair. Seems like it is good to have separate lineups, beyond the main good one. That might be alright or better if further optimized along those lines but there could be issues. More to look at, including what they close with.

Kings main starting lineup the last 21 games was a hideous -19pts / 100p. They did go 11-10, but it was because of bench lineups. Almost all the bigger minute positives had Hield. One way to do it. Probably better ways, but would have to try & prove.

That lineup wasn't used in every game. An apparently survivable -6/100 in wins. A ridiculous -45 in losses. Probably need to yank quick when it begins to go off rails.

I'd want a different start.

Conley - Mitchell - Ingles is the worst bigger minute trio for each.

-8pts / 100p for season. But Snyder used it more frequently and for more minutes per game late in season despite it slipping to -13. Stubborn or ignorant of this information? Neither is an acceptable answer.

Rubio Booker Oubre Bridges Ayton is plus 19pts / 100p but used just 3.5 minutes per game for season.

Rubio Booker Oubre Bridges works with Baynes and others. Get down to just 3 of those 4 and things are hit n miss.

Should probably go far heavier with lineups with those 4 together. And with 3 in lineups that work.