As mentioned in Thunder thread, there are only 29 lineups averaging over 3.5 minutes per game, less than 1 per team. 5 top contenders have biggest minute lineups that are +9 to +17. The 5 are Bucks, Celtics, Clippers, Nuggets and Timberwolves. Among the playoff bound failing to have a qualifer performer at the +9 or above level are the Thunder, Cavs, Knicks, Sixers, Suns, Kings, Magic, Pelicans, Mavs, Pacers, Pelicans, Lakers and Warriors.
On average teams have a bit less than 7 lineups tested to a low bar of 50 minutes or above for season. Probably less than 2% of lineups are non-dink (dink being less than 1 minute per game for season).
Thunder are at 197 for super, super dink lineups of under 10 minutes for season. Warriors at 360. Pistons, 531 such.
The relative lack of concentration at the top and massive use of near dink, dink and super dink lineups remains probably the greatest inefficiency and analytic failure in the NBA. But it persists, with little visible sign of attention or concern or efforts to change the pattern from Coaches, analytics shops, GMs and Owners. Some fans comment on it ocassionally, the media rarely.
LIneup trends around the league
Re: League wide lineup trends
How can it be demonstrated that frequent and varied substitutions are inefficient and a relative 'failure'?
You could have a steady starting 5, and a 2nd 5, and a quasi-3rd 5. But when any player is unavailable, that plan is wrecked.
At what point would you banish a given lineup? At 50 minutes below -10? If sooner, more dinks; if later, you're just losing.
If a coach wants to see all his players in action and in different combinations, he can't give them all 50+ minutes.
I've always thought one sign of an insecure coach is that he sticks too long with a lineup that has done well. A great coach knows to sub a guy out before it's overdue.
A coach tries to stay a step ahead of the opponent.
You could have a steady starting 5, and a 2nd 5, and a quasi-3rd 5. But when any player is unavailable, that plan is wrecked.
At what point would you banish a given lineup? At 50 minutes below -10? If sooner, more dinks; if later, you're just losing.
If a coach wants to see all his players in action and in different combinations, he can't give them all 50+ minutes.
I've always thought one sign of an insecure coach is that he sticks too long with a lineup that has done well. A great coach knows to sub a guy out before it's overdue.
A coach tries to stay a step ahead of the opponent.
Re: League wide lineup trends
"How can it be demonstrated that frequent and varied substitutions are inefficient and a relative 'failure'?"
It was shown that bigger minute lineups have a better average performance by Eli Witus perhaps a decade ago and I've made notes about it repeatedly over time. Maybe I'll do it again, but maybe folks will someday give proper credit to and concentration on bigger minute, more proven lineups in contrast to the dink lineups regime.
"If a coach wants to see all his players in action and in different combinations, he can't give them all 50+ minutes."
How much does a Coach and players really learn playing in 500-1000 lineups in a season compared to 200 or 50? At 1000 lineups, the average use would be a ridiculous 5 minutes. What was really learned?
At 500 lineups, the average use would be a ridiculous 10 minutes. What was really learned about that lineup's expected, future effectiveness?
At 200 lineups, the average use would be a 25 minutes in a full season. Still not enough on average to judge. Need to limit to 100 to get to 50 minutes on average. To get even some to the more statistically significant level to 200 or 400 minute test of the most important, you got to cut way down on dink lineups.
You CAN'T test everything and everything doesn't NEED to be tested and seen and experienced AS MUCH AS A HANDFUL OR A COUPLE OF HANDFULS ARE NEEDED AND NEED TO BE MORE FULLY TESTED FOR THE PLAYOFFS, as compared to the current practice of testing just 1-2 lineups over 100 minutes.
If a Coach can't use insight and experience to envision 5 -10 likely most important lineups for playoffs, what good is he?
If he can't envision most important 25-100 or 200 important lineups, enough to cover most or all of the range of contexts & circumstances, what good is he? Do you want a coach to use 500-1000 lineups or 200 where amount of test / possible learning on average is inverse to number of lineups used? I know my answer.
It was shown that bigger minute lineups have a better average performance by Eli Witus perhaps a decade ago and I've made notes about it repeatedly over time. Maybe I'll do it again, but maybe folks will someday give proper credit to and concentration on bigger minute, more proven lineups in contrast to the dink lineups regime.
"If a coach wants to see all his players in action and in different combinations, he can't give them all 50+ minutes."
How much does a Coach and players really learn playing in 500-1000 lineups in a season compared to 200 or 50? At 1000 lineups, the average use would be a ridiculous 5 minutes. What was really learned?
At 500 lineups, the average use would be a ridiculous 10 minutes. What was really learned about that lineup's expected, future effectiveness?
At 200 lineups, the average use would be a 25 minutes in a full season. Still not enough on average to judge. Need to limit to 100 to get to 50 minutes on average. To get even some to the more statistically significant level to 200 or 400 minute test of the most important, you got to cut way down on dink lineups.
You CAN'T test everything and everything doesn't NEED to be tested and seen and experienced AS MUCH AS A HANDFUL OR A COUPLE OF HANDFULS ARE NEEDED AND NEED TO BE MORE FULLY TESTED FOR THE PLAYOFFS, as compared to the current practice of testing just 1-2 lineups over 100 minutes.
If a Coach can't use insight and experience to envision 5 -10 likely most important lineups for playoffs, what good is he?
If he can't envision most important 25-100 or 200 important lineups, enough to cover most or all of the range of contexts & circumstances, what good is he? Do you want a coach to use 500-1000 lineups or 200 where amount of test / possible learning on average is inverse to number of lineups used? I know my answer.
Re: League wide lineup trends
OKC Thunder:
Average performance of 3 most used, only non dink lineups, +9.6 per 48 minutes.
Avg for 437 dink lineups, +7.2. 33% less than for the non-dink.
Do you want to win by 33% more per minute up to a feasible max of time and court effectiveness? I do.
Average performance of 3 most used, only non dink lineups, +9.6 per 48 minutes.
Avg for 437 dink lineups, +7.2. 33% less than for the non-dink.
Do you want to win by 33% more per minute up to a feasible max of time and court effectiveness? I do.
Re: League wide lineup trends
Crow
Wed Jul 19, 2023 3:29 pm
"Imo, the ideal distribution of minutes to lineups for a team trying to be it's best would be something like:
1 lineup of 800-1000 minutes (your best performer)
2 lineups of 400 minutes (your best alternatives)
4-6 of 200 minutes
maybe 5 more at or near 100 minutes
then about 500-1000 minutes of situational calls.
Last season, only 3 teams had a lineup used 8 minutes per game for season. Only 12 had a lineup even over 400 minutes as a top lineup. Instead of 6-8 lineups at or above 200 minutes as in my model, the actual average was 0.7. Instead of 12-14 lineups over 100 minutes, the actual average was 3.3.
My model might not be exactly right for every team, especially if heavily affected by injuries or change, but the actual average is way short of desirable for every team. There would be a middle ground, giving some priority to development of / experimentation with young or new, but that middle ground of concentration would still be higher to far higher than the current norm."
Wed Jul 19, 2023 3:29 pm
"Imo, the ideal distribution of minutes to lineups for a team trying to be it's best would be something like:
1 lineup of 800-1000 minutes (your best performer)
2 lineups of 400 minutes (your best alternatives)
4-6 of 200 minutes
maybe 5 more at or near 100 minutes
then about 500-1000 minutes of situational calls.
Last season, only 3 teams had a lineup used 8 minutes per game for season. Only 12 had a lineup even over 400 minutes as a top lineup. Instead of 6-8 lineups at or above 200 minutes as in my model, the actual average was 0.7. Instead of 12-14 lineups over 100 minutes, the actual average was 3.3.
My model might not be exactly right for every team, especially if heavily affected by injuries or change, but the actual average is way short of desirable for every team. There would be a middle ground, giving some priority to development of / experimentation with young or new, but that middle ground of concentration would still be higher to far higher than the current norm."
Re: League wide lineup trends
Wed Jul 19, 2023 5:46 pm
"The simplest way to improve is probably just to take the best coach chosen lineups and increase their utilization greatly...
Close to 60% of the most used 100 lineups have been used less than 200 minutes, Any over +5 or +10 should probably see 50-100% increase in utilization (in perceived favorable circumstances based on split analysis) for awhile and see what happens. There were almost 20 cases in league were a +7.5 lineup over 100 minutes failed to get 200 minutes. Warriors 1, Celtics 1, Heat 3, Bucks 2, Griz 2, Suns 1, Lakers 1, Clippers 2... Nuggets 0"
"The simplest way to improve is probably just to take the best coach chosen lineups and increase their utilization greatly...
Close to 60% of the most used 100 lineups have been used less than 200 minutes, Any over +5 or +10 should probably see 50-100% increase in utilization (in perceived favorable circumstances based on split analysis) for awhile and see what happens. There were almost 20 cases in league were a +7.5 lineup over 100 minutes failed to get 200 minutes. Warriors 1, Celtics 1, Heat 3, Bucks 2, Griz 2, Suns 1, Lakers 1, Clippers 2... Nuggets 0"
Re: League wide lineup trends
Wed Jul 19, 2023 6:24 pm
"Less than 62% of player pairs over 500 minutes were positive on raw +/-. The average team had 24 such pairs with 9 non-positive. How many of them were / will be directly reviewed and how many will be diminished or eliminated?"
Diminish or eliminate the worst of these, perhaps on a rolling basis.
"Less than 62% of player pairs over 500 minutes were positive on raw +/-. The average team had 24 such pairs with 9 non-positive. How many of them were / will be directly reviewed and how many will be diminished or eliminated?"
Diminish or eliminate the worst of these, perhaps on a rolling basis.
Re: League wide lineup trends
Thu Jul 20, 2023 9:08 am
"I said elsewhere recently that 95.5% of the champion's net playoff margin over last 5 years came from their 5 biggest lineups. They didn't get outcoached by playing familiar.
More generally 25 of the 31 biggest minute lineups in the last playoffs were positive. Over 80%."
"I said elsewhere recently that 95.5% of the champion's net playoff margin over last 5 years came from their 5 biggest lineups. They didn't get outcoached by playing familiar.
More generally 25 of the 31 biggest minute lineups in the last playoffs were positive. Over 80%."
Re: League wide lineup trends
"The response to small sample experience is to ... try to beat or dance ahead of chance, or at least the last negative experience in short term memory, never or almost never having explored life under much larger sample sizes."
Playoff behavior might be more in the moment abrupt than regular season behavior but abrupt behavior is often overly abrupt and based on just a few bad possessions. If you really think a matchup is irrevocably doomed, change; but you shouldn't be in that many of those situations if you are playing lineups with
good track records in diverse situations as opposed to heat of moment, seat of pants, untested or lightly tested dink lineups seeking momentary advantage or recovery.
A successful Chessmaster if probably going to play the most tried, best proven move most of the time. Try something less tried, less proven? Maybe sometimes if really well thought out, my insight superior to yours, but not just off the cuff, looks good at first glance.
Playoff behavior might be more in the moment abrupt than regular season behavior but abrupt behavior is often overly abrupt and based on just a few bad possessions. If you really think a matchup is irrevocably doomed, change; but you shouldn't be in that many of those situations if you are playing lineups with
good track records in diverse situations as opposed to heat of moment, seat of pants, untested or lightly tested dink lineups seeking momentary advantage or recovery.
A successful Chessmaster if probably going to play the most tried, best proven move most of the time. Try something less tried, less proven? Maybe sometimes if really well thought out, my insight superior to yours, but not just off the cuff, looks good at first glance.
Re: League wide lineup trends
Mike G: "At what point would you banish a given lineup? At 50 minutes below -10? If sooner, more dinks; if later, you're just losing."
If the response to abandoning a dink lineup is another dink, you stay in the same chaos.
The answers are to limit lineup trial quantity from the start and to replace non-early successes with more time with the better tested, better performers as long as they continue to perform well. Maybe limit their exposure to possibly bad contexts but that should be tested some and not be knee-jerk reductions or cessations based on a few minutes in one stint.
I'll have to think more about when to stop an initial lineup trial but your suggestion of 50 minutes at -10 would be one level. Might be sooner if much worse than -10 or if much worse than expected / desired by stats and / or eye test. Lineup management should have a scientific core; but given the limited total time available, it will require artistic judgment calls about how to maximize testing time and results.
When to curtail or stop a lineup with meh or bad early results will partially depend on how much you believed in it apriori. Thunder stuck with SGA Giddey Dort when early results were quite bad. It got better in longer test and with better finishing pieces and better management of what they were doing in it (from Giddey as PG to back firmly to SGA; from Dort shooting a lot / too much to quite a bit less).
Raw data is of course inferior to adjusted but raw is more quickly available for outside commentary. Working on the inside, yeah, adjust and know the strategies used in lineups and all the play calls, etc.
If the response to abandoning a dink lineup is another dink, you stay in the same chaos.
The answers are to limit lineup trial quantity from the start and to replace non-early successes with more time with the better tested, better performers as long as they continue to perform well. Maybe limit their exposure to possibly bad contexts but that should be tested some and not be knee-jerk reductions or cessations based on a few minutes in one stint.
I'll have to think more about when to stop an initial lineup trial but your suggestion of 50 minutes at -10 would be one level. Might be sooner if much worse than -10 or if much worse than expected / desired by stats and / or eye test. Lineup management should have a scientific core; but given the limited total time available, it will require artistic judgment calls about how to maximize testing time and results.
When to curtail or stop a lineup with meh or bad early results will partially depend on how much you believed in it apriori. Thunder stuck with SGA Giddey Dort when early results were quite bad. It got better in longer test and with better finishing pieces and better management of what they were doing in it (from Giddey as PG to back firmly to SGA; from Dort shooting a lot / too much to quite a bit less).
Raw data is of course inferior to adjusted but raw is more quickly available for outside commentary. Working on the inside, yeah, adjust and know the strategies used in lineups and all the play calls, etc.
Re: League wide lineup trends
Celtics non-dink lineups on average only 14% better than average for dink lineups but I'd jump on anything offering 14% better results. Or 33% or even a few %. This would especially be true for non-dominant teams who could greatly benefit from every extra point in close games, to keep them close longer, to give more confidence to fight down the stretch.
Re: League wide lineup trends
Trying to be 1 or 2 steps ahead of opponent machine gun coaching is one approach; but is it better than relying on players in their best proven on average lineups? It depends how convinced of and committed you are to your own wizardry compared to your core player talent and fit.
How many titles were won first and foremost by coaching wizardry over core player talent and fit?
If trying to be 1-2 steps ahead worked, dink lineups would perform better than they do, perhaps better than the longer tried. That might be rarely true for specific Coaches and teams but probably not in general.
How many titles were won first and foremost by coaching wizardry over core player talent and fit?
If trying to be 1-2 steps ahead worked, dink lineups would perform better than they do, perhaps better than the longer tried. That might be rarely true for specific Coaches and teams but probably not in general.
Re: League wide lineup trends
Nuggets 2 most used and only lineups over 1.5 minutes per game, +14.1 pts / 100p. 3 near dink and 289 actual dink or super dink lineups combined average,-1.7 pts / 100p.
Spurs 3 most used and only lineups over 1.5 minutes per game, +7.5 pts / 100p. 2 near dink and 407 actual dink or super dink lineups combined average,-11.3 pts / 100p.
Sixers 2 most used and only lineups over 1.7 minutes per game, +19.0 pts / 100p. 2 near dink and 519 actual dink or super dink lineups combined average,+0.5 pts / 100p.
The difference can be huge.
Best players can only play so much, but best lineups can often be played more to far, far more.
Calculating expected performance would take more work but could be done. The difference will be some unknown mix of synergy and randomness.
But either you pursue the strong performers hard or you don't. Which is a better guess? I am going with the strongest performers that have gotten the most test until they stop performing long enough for that impression to replace the earlier one.
Spurs 3 most used and only lineups over 1.5 minutes per game, +7.5 pts / 100p. 2 near dink and 407 actual dink or super dink lineups combined average,-11.3 pts / 100p.
Sixers 2 most used and only lineups over 1.7 minutes per game, +19.0 pts / 100p. 2 near dink and 519 actual dink or super dink lineups combined average,+0.5 pts / 100p.
The difference can be huge.
Best players can only play so much, but best lineups can often be played more to far, far more.
Calculating expected performance would take more work but could be done. The difference will be some unknown mix of synergy and randomness.
But either you pursue the strong performers hard or you don't. Which is a better guess? I am going with the strongest performers that have gotten the most test until they stop performing long enough for that impression to replace the earlier one.
Re: League wide lineup trends
"You could have a steady starting 5, and a 2nd 5, and a quasi-3rd 5. But when any player is unavailable, that plan is wrecked."
You could potentially have completely separate lineups/ shifts be optimal, but more likely it could be something like:
Starters
Starters with one sub
Starters with 2 subs
Starters with several
more different sets of 2 subs
Starters with a different sub
Starters with a still different sub
Starters
You could potentially have completely separate lineups/ shifts be optimal, but more likely it could be something like:
Starters
Starters with one sub
Starters with 2 subs
Starters with several
more different sets of 2 subs
Starters with a different sub
Starters with a still different sub
Starters
Re: League wide lineup trends
Using a last game filter, I find that the average team in mid-March used 14.5 lineups. That translates to 3.3 minutes per lineup or almost exactly 200 seconds.
The longest runs was 25 minutes (Cavs). 8 at or over 15 minutes. There are an average of a bit over 1 10+ minute lineup per team. 2 over 6 minutes, 4-5 over 3 minutes. 10 over 1 minute.
Dinks are created via length of game use but moreso probably by game use infrequency.
Slight less than half of the 10+ minute lineups were positive but most strongly so. The 13 big minute positives won by an average of 7 raw points. The 17 negatives lost by slightly more. The other 413 lineups won slightly overall but on average by only +0.1 pts each.
Lineups with 3+minutes had almost 50% rate of being positive. Lineups with less time had modestly over a 25% rate of being positive.
One game, may not be exactly centered but some quick impressions / things to learn, if interested.
The average team's lineups involved about 42 player pairs. About 42% were positive. 30.5 got 5+ minutes and about 45% of those were positive. 20 got 10+ minutes and about 47% of those were positive. Remove the bigger minute sub-group from the larger population and the differences in sub-group performance would be greater.
Can glean some trends related to degree of concentration.
Of course, more could be done.
The longest runs was 25 minutes (Cavs). 8 at or over 15 minutes. There are an average of a bit over 1 10+ minute lineup per team. 2 over 6 minutes, 4-5 over 3 minutes. 10 over 1 minute.
Dinks are created via length of game use but moreso probably by game use infrequency.
Slight less than half of the 10+ minute lineups were positive but most strongly so. The 13 big minute positives won by an average of 7 raw points. The 17 negatives lost by slightly more. The other 413 lineups won slightly overall but on average by only +0.1 pts each.
Lineups with 3+minutes had almost 50% rate of being positive. Lineups with less time had modestly over a 25% rate of being positive.
One game, may not be exactly centered but some quick impressions / things to learn, if interested.
The average team's lineups involved about 42 player pairs. About 42% were positive. 30.5 got 5+ minutes and about 45% of those were positive. 20 got 10+ minutes and about 47% of those were positive. Remove the bigger minute sub-group from the larger population and the differences in sub-group performance would be greater.
Can glean some trends related to degree of concentration.
Of course, more could be done.