Celtics Offensive and defensive efficiencies both up (compared to earlier).
But 238 lineups used already and only 7 used over 20 minutes total? This is what Brad Stevens and lineup guru Drew Cannon have gone with? Rondo gone so only 3 of the lineups are still possible and positive. None used over a paltry 30 minutes yet. Averaging about 5 minutes a game, they won't end up with a very meaningful season test that way to use for future planning. Guess your best and test them over 500 minutes. Or do the coaching gametime match-up thing and have 400-600 rows of data at end of year with virtually nothing even semi-solid to look at. But with all those draft picks coming, maybe testing lineups doesn't really matter for a few years. Except for practice and learning.
Checking in on Celtics lineup usage (and other teams)
Re: Checking in on the Celtics
By RPM estimates, the Celtics have good in the paint players and nearly all bad perimeter players. Smart isn't great on boxscore but his neutral RPM this early is encouraging.
Starting lineup with Turner working so far but his overall RPM is quite bad. Will both last, will "proper usage" raise the RPM or sample size catch up with the starting lineup?
Starting lineup with Turner working so far but his overall RPM is quite bad. Will both last, will "proper usage" raise the RPM or sample size catch up with the starting lineup?
Re: Checking in on the Celtics
I assume the starting lineup of Turner-Bradley-Green-Sullinger-Zeller was chosen to mimic the same lineup with Rondo that performed well. This new version worked the first few games but on average hasn't since. Looking here at trios http://www.basketball-reference.com/pla ... rder_by=mp
you can see it contains several really bad performing trios and none of the strong ones. Based on strong trios, the best lineup might have Smart-Turner-Thornton or Green-Bass-Olynyk. And for the moment, these are the 2 best lineups used over 20 minutes. http://www.basketball-reference.com/pla ... rder_by=mp But in the chaos of the last 10 games, 97 lineups were used, but neither of these, unless I missed them in the litter. Maybe this is small sample mirage too but at least it has larger sample support at the trio level.
Overall RPM argues for Sullinger. But the current starting lineup isn't looking like the right specific lineup at the moment. Experiment but find something that works and has depth of support for working.
If you aren't really focused on making the playoffs this season, you might not have interest in Bass lineups. Or Green lineups. Or Thornton or probably Turner either. But choose a focus. A little of everything produces little information with any substance. If the team's best guess is the current starters, then continue to test it. Really there is time to test a few and compare. But not hundreds.
you can see it contains several really bad performing trios and none of the strong ones. Based on strong trios, the best lineup might have Smart-Turner-Thornton or Green-Bass-Olynyk. And for the moment, these are the 2 best lineups used over 20 minutes. http://www.basketball-reference.com/pla ... rder_by=mp But in the chaos of the last 10 games, 97 lineups were used, but neither of these, unless I missed them in the litter. Maybe this is small sample mirage too but at least it has larger sample support at the trio level.
Overall RPM argues for Sullinger. But the current starting lineup isn't looking like the right specific lineup at the moment. Experiment but find something that works and has depth of support for working.
If you aren't really focused on making the playoffs this season, you might not have interest in Bass lineups. Or Green lineups. Or Thornton or probably Turner either. But choose a focus. A little of everything produces little information with any substance. If the team's best guess is the current starters, then continue to test it. Really there is time to test a few and compare. But not hundreds.
Re: Checking in on the Celtics
B. Bass | A. Bradley | M. Smart | E. Turner | T. Zeller the by far most used lineup at season end was a pretty bad -6.9 per 100 possessions. Do they bag that or roll it out as the starting lineup next season? They are high on Crowder but two of his top 3 most lineups were horrendous in small sample usage. B. Bass | A. Bradley | J. Crowder | E. Turner | T. Zeller was +3, try that more? B. Bass | K. Olynyk | M. Smart | M. Thornton | E. Turner was dynamite... in all of 43 minutes of trial. Not a single lineup over 26 minutes with Sullinger that was positive is still possible. What positive was really learned about full lineups this season that can be applied directly to next season? Not much.
Re: Checking in on the Celtics
Stevens lineup management in playoffs:
96 lineups used in 192 minutes.
?
96 lineups used in 192 minutes! An average of 2 minutes in the series for each careful, tactical decision.
Only 9 used a full minute per game.
Only 3 used 2 minutes per game.
None over 5 minutes per game, only one much over 3 minutes per game. And it was negative, pretty bad per 100 possessions, though it got only 40-50% of the way to that first 100.
Two thirds of lineups were neutral or negative.
Lost on net 4 factors in 14 of 16 game cases. Hardly any net improvement from game 1 thru 4. All losses of 8 to 11.
This is how they chose do it? Not the way I'd go.
CLE? 99 lineups used. A bit less than 50% positive. GSW? 108 lineups used, under 40% positive. So Celtics used about the same number of lineups but in a bit more than 1/4th the time. Is there really a need for 100 lineups in the playoffs? I doubt there is really a need for that many and think it might be optimal to keep it under or well under 50 but I'll check some more history to see what it shows.
96 lineups used in 192 minutes.
?
96 lineups used in 192 minutes! An average of 2 minutes in the series for each careful, tactical decision.
Only 9 used a full minute per game.
Only 3 used 2 minutes per game.
None over 5 minutes per game, only one much over 3 minutes per game. And it was negative, pretty bad per 100 possessions, though it got only 40-50% of the way to that first 100.
Two thirds of lineups were neutral or negative.
Lost on net 4 factors in 14 of 16 game cases. Hardly any net improvement from game 1 thru 4. All losses of 8 to 11.
This is how they chose do it? Not the way I'd go.
CLE? 99 lineups used. A bit less than 50% positive. GSW? 108 lineups used, under 40% positive. So Celtics used about the same number of lineups but in a bit more than 1/4th the time. Is there really a need for 100 lineups in the playoffs? I doubt there is really a need for that many and think it might be optimal to keep it under or well under 50 but I'll check some more history to see what it shows.
Re: Checking in on the Celtics
For last 5 champs, the average is 106 lineups, though in more games generally. High was Pop last season at 145. Carlisle with the low of 77. (He only used 28 when they were swept in first round of following season so below 50 has happened. Didn't work that time.) In every case less than 50% of the lineups chosen by the title winner were positive. The average was right at 40% or a little under. 2011 Mavs were under 30%.
Call it experimentation? Maybe but how many of the lineups tried in playoffs deserved to be tried or even hinted of strong potential? I thought and still think the regular season is the time for experimentation, preferably very disciplined, selective, strategic experimentation, not 350-600 lineup experimentation. 70-85% of those don't even make it to the 100 lineup playoff set. 95% of them got so little testing that not even the most earnest, thoughtful lineup expert could claim he or anyone knows anything solid or semi-solid about them.
I'd focus almost all the time on the top 20-50. If you are going to be basically blind on everything else, at least try to focus on and maybe learn something fuzzy about 20 or 50. Or to be more realistic, 5-8 main lineups. You could do a playoff rotation in 5-8 lineups, if you want to, if you designed for it, if it looks good. Of course depart from a smaller than normal set if situation calls for it but it seems like coaches are going micro-tactical far more than they are presenting a strategic package. Maybe my vision is out of touch with the moment to moment reality but it is looks like chaos. If they were such great micro-tactical geniuses why are the titlewinners, the best of the them, winning under 50%, under 40% with these micro-lineups? They are winning with their bigger lineups generally I assume. The small lineups are not really helping much or even hurting. Fortunately they are not helping or hurting the other guy as much or worse.
Call it experimentation? Maybe but how many of the lineups tried in playoffs deserved to be tried or even hinted of strong potential? I thought and still think the regular season is the time for experimentation, preferably very disciplined, selective, strategic experimentation, not 350-600 lineup experimentation. 70-85% of those don't even make it to the 100 lineup playoff set. 95% of them got so little testing that not even the most earnest, thoughtful lineup expert could claim he or anyone knows anything solid or semi-solid about them.
I'd focus almost all the time on the top 20-50. If you are going to be basically blind on everything else, at least try to focus on and maybe learn something fuzzy about 20 or 50. Or to be more realistic, 5-8 main lineups. You could do a playoff rotation in 5-8 lineups, if you want to, if you designed for it, if it looks good. Of course depart from a smaller than normal set if situation calls for it but it seems like coaches are going micro-tactical far more than they are presenting a strategic package. Maybe my vision is out of touch with the moment to moment reality but it is looks like chaos. If they were such great micro-tactical geniuses why are the titlewinners, the best of the them, winning under 50%, under 40% with these micro-lineups? They are winning with their bigger lineups generally I assume. The small lineups are not really helping much or even hurting. Fortunately they are not helping or hurting the other guy as much or worse.
Re: Checking in on the Celtics
2001 Lakers, 72 lineups used. 85% of net point margin from top 5 performing lineups.
2004 Pistons, 157 lineups. Top five achieved 135% of net point margin. The other 152 Larry Brown decided to use based of his micro-tactical judgment lost by 55 points.
2007 Spurs top five lineups contributed exactly 100% of net margin. Other 99 lineups combined added exactly nothing.
2008 Celtics 129 lineups in playoffs. Top 5 achieve 110% of net margin. Doc's other 124 choices lost by 12.
2011 Mavs, top 5 contributed 175% of margin. Other 72 lose by 95 points.
2012 Heat, top 5 contributed 85% of margin. Other 119 add about a point per game to the margin.
2014 Spurs. The least concentrated, most lineup happy recent title winner. Still 75% of their point margin came from 5 lineups. 25% from the other 140.
Which is more important, knowing and playing your top 5 lineups at least much as your opponent or trying to win within the vast set of micro-tactical lineups? Based on what I've found so far (haven't checked all but no real big picture outlier so far), I know my answer. I'd played my top 5 to their max or optimal cutoff. I see no coach yet that I'd rely on to win the micro-tactical lineup battle outside the top 5 and win the title that way.
Stevens in the regular season? Top 5 lineups were plus 141. The other 840 lineups?? Minus 128. But the media lavish praise on a few dramatically visual examples of late game tactical genius. Only 32% of lineups used were positive for the season, playing one of the weakest east schedules.
Pop played 747 rs lineups. And genius that he is... he won for the season on 1/3rd of them. Why would a coach or GM vigorously pursue or allow such strategy? You might get away with it if the other side is lighter of top lineup usage and / or similar or worse on non top 5 lineup productivity but why not be more concentrated / selective / results driven?
Scott Brooks had to deal with injuries and change, but he didn't have to try 630 lineups, of which he won with 35% of them. 7 of the ten best performing lineups were without KD and probably available for 30 plus games generally... and yet despite being best performing, none of these got used over 63 minutes for the season.
2004 Pistons, 157 lineups. Top five achieved 135% of net point margin. The other 152 Larry Brown decided to use based of his micro-tactical judgment lost by 55 points.
2007 Spurs top five lineups contributed exactly 100% of net margin. Other 99 lineups combined added exactly nothing.
2008 Celtics 129 lineups in playoffs. Top 5 achieve 110% of net margin. Doc's other 124 choices lost by 12.
2011 Mavs, top 5 contributed 175% of margin. Other 72 lose by 95 points.
2012 Heat, top 5 contributed 85% of margin. Other 119 add about a point per game to the margin.
2014 Spurs. The least concentrated, most lineup happy recent title winner. Still 75% of their point margin came from 5 lineups. 25% from the other 140.
Which is more important, knowing and playing your top 5 lineups at least much as your opponent or trying to win within the vast set of micro-tactical lineups? Based on what I've found so far (haven't checked all but no real big picture outlier so far), I know my answer. I'd played my top 5 to their max or optimal cutoff. I see no coach yet that I'd rely on to win the micro-tactical lineup battle outside the top 5 and win the title that way.
Stevens in the regular season? Top 5 lineups were plus 141. The other 840 lineups?? Minus 128. But the media lavish praise on a few dramatically visual examples of late game tactical genius. Only 32% of lineups used were positive for the season, playing one of the weakest east schedules.
Pop played 747 rs lineups. And genius that he is... he won for the season on 1/3rd of them. Why would a coach or GM vigorously pursue or allow such strategy? You might get away with it if the other side is lighter of top lineup usage and / or similar or worse on non top 5 lineup productivity but why not be more concentrated / selective / results driven?
Scott Brooks had to deal with injuries and change, but he didn't have to try 630 lineups, of which he won with 35% of them. 7 of the ten best performing lineups were without KD and probably available for 30 plus games generally... and yet despite being best performing, none of these got used over 63 minutes for the season.
Re: Checking in on the Celtics
3 of the 4 conference finalists had one of the top 4 regular season lineups by net point margin contribution. The Warriors had 4 of the top 15. (Cavs 2, nobody else more than one, but all the first round winners but the Bulls and Griz had a top 15. Griz 17th. Bulls 40th.)
Re: Checking in on Celtics lineup usage (and other teams)
One way to plan lineup set. Ask coach to select 8 lineups under the constraint that is all they get to choose from for game / season (assuming no injuries). After that, and only after that, ask them to pick 12 and only 12 alternatives to deal with expected and desired matchup situations.
After that, give them a list of 1000 other possible lineups and ask for a specific reason or situation why and when that lineup is expected to be better than the 20 already selected... and yet was not selected in the 20. Do this for all 1000 or more. All of them, specific, prior analysis. Could be 1-2 sentences but do all of them and review all of them and decide whether the rationale is reasonable, necessary and sufficient or not. And then review again and pick a very special situation third tier of only the next best and helpful 30-80. And deep 6 or at least deep 3 all of the rest barring a really unexpected emergency.
Of course need will arise during actual season to use more due to injury and poor performance and so-called garbage time (could be renamed and repurposed as important future research time) but a team should ideally know its top 20 lineup set and then a top 50-100 and not casually depart from it unless there is good reason to believe it will be better than their best prior thinking. This could essentially mean having 2.1-2.2 options at every position and playing every combination. If you pass on playing some of the combinations you could have fractionally more options at some or all positions. It is that somehow too restrictive for a PLAN? Playing 845 lineups is essentially having about 3.7 - 3.8 options at every position, which is possible in a heavy trade season but then the key is playing every combination, which is probably not necessary or helpful.
Coaches probably don't want to go thru this, but shouldn't they? How many of them do something similar to this, on own initiative or on request? As I said this is one way to approach it. Seems worthwhile to me. In contrast to maybe having an idea of the top 5-8 and then winging the rest in the moment without explicit prior analysis or prioritization.
If you gave yourself a second 100 to deal with injuries, serious future oriented research / player development and I am too tired or frustrated for that garbage time then you are at 200 and less than half of most or nearly all teams lineup counts and less than 1/4 this season's Celtics. I'd interested to know the all-time high and low counts but haven't looked yet.
After that, give them a list of 1000 other possible lineups and ask for a specific reason or situation why and when that lineup is expected to be better than the 20 already selected... and yet was not selected in the 20. Do this for all 1000 or more. All of them, specific, prior analysis. Could be 1-2 sentences but do all of them and review all of them and decide whether the rationale is reasonable, necessary and sufficient or not. And then review again and pick a very special situation third tier of only the next best and helpful 30-80. And deep 6 or at least deep 3 all of the rest barring a really unexpected emergency.
Of course need will arise during actual season to use more due to injury and poor performance and so-called garbage time (could be renamed and repurposed as important future research time) but a team should ideally know its top 20 lineup set and then a top 50-100 and not casually depart from it unless there is good reason to believe it will be better than their best prior thinking. This could essentially mean having 2.1-2.2 options at every position and playing every combination. If you pass on playing some of the combinations you could have fractionally more options at some or all positions. It is that somehow too restrictive for a PLAN? Playing 845 lineups is essentially having about 3.7 - 3.8 options at every position, which is possible in a heavy trade season but then the key is playing every combination, which is probably not necessary or helpful.
Coaches probably don't want to go thru this, but shouldn't they? How many of them do something similar to this, on own initiative or on request? As I said this is one way to approach it. Seems worthwhile to me. In contrast to maybe having an idea of the top 5-8 and then winging the rest in the moment without explicit prior analysis or prioritization.
If you gave yourself a second 100 to deal with injuries, serious future oriented research / player development and I am too tired or frustrated for that garbage time then you are at 200 and less than half of most or nearly all teams lineup counts and less than 1/4 this season's Celtics. I'd interested to know the all-time high and low counts but haven't looked yet.
Re: Checking in on Celtics lineup usage (and other teams)
Another line of thinking would be to say, if you used a lineup less than 5 minutes this entire season, then you didn't want or think you needed it much and you probably didn't really need it. By this approach Stevens probably didn't need almost 700 of the lineups he used. Half of all his lineups got less than 2 minutes. With only about 4,000 regular minutes that is chewing up a lot of available time on super-dink lineups, most of which performed poorly.
Warriors had mostly good health and good results with 342 lineups. Still only about 130 got used more than 5 minutes for the season. Nearly 2/3rds played less than half a quarter and weren't that necessary.
This isn't really about tidiness. It is about optimizing testing and deployment of the apparent best or better lineups, the ones perceived wanted and needed more than a few minutes in a season and might get used in the playoffs.
Warriors had mostly good health and good results with 342 lineups. Still only about 130 got used more than 5 minutes for the season. Nearly 2/3rds played less than half a quarter and weren't that necessary.
This isn't really about tidiness. It is about optimizing testing and deployment of the apparent best or better lineups, the ones perceived wanted and needed more than a few minutes in a season and might get used in the playoffs.
Re: Checking in on Celtics lineup usage (and other teams)
Lineup entropy might be of some use here in understanding how coaches use/vary lineups. A while back, I found some evidence that low lineup entropy in one season (more consistency in lineups) led to early season success the following year.
http://www.basketballanalyticsbook.com/ ... n-success/
http://www.basketballanalyticsbook.com/ ... n-success/
Re: Checking in on Celtics lineup usage (and other teams)
Lineups used over 500 minutes in 2014-15 season: 5. The fewest for which season data is available at BR (15 yrs worth). Typically it is 10+ and has been as high as 19. 2000-1's 6 was the only season close to as low as 2014-15. Is the trend to rest stars contributing? I'd rather reduce their usage in dink lineups than the most used ones.
Is low usage of the best lineups possibly one of he biggest and most neglected issues since the sluggish historical ramp up in the use of the 3 pt shot until fairly recently?
15 of the top 18 most used lineups last season were positive, by an average of almost plus 10 pts per 100 possessions. Availability may be an issue, given games used; but only 3 of these 18 got used on average a full 15 minutes per game when available. I can understand the pull to divide stars amongst more or all the minutes but do teams really know if this strategy is paying off or not net? It better be paying back handsomely to decline the possibility of more minutes from top lineups producing at this average rate.
Is low usage of the best lineups possibly one of he biggest and most neglected issues since the sluggish historical ramp up in the use of the 3 pt shot until fairly recently?
15 of the top 18 most used lineups last season were positive, by an average of almost plus 10 pts per 100 possessions. Availability may be an issue, given games used; but only 3 of these 18 got used on average a full 15 minutes per game when available. I can understand the pull to divide stars amongst more or all the minutes but do teams really know if this strategy is paying off or not net? It better be paying back handsomely to decline the possibility of more minutes from top lineups producing at this average rate.
Re: Checking in on Celtics lineup usage (and other teams)
A few more notes on the subject:
Most used Celtics lineup last season? Was 4th worst lineup in league used over 250 minutes.
Only Celtics lineup used over 50 minutes still possible? -16 per 100 possessions.
Odds it gets used a lot (at least relative to others) next season? Pretty good.
Only 4 teams used fewer lineups over 50 minutes last season. Only two more than barely.
Most used Celtics lineup last season? Was 4th worst lineup in league used over 250 minutes.
Only Celtics lineup used over 50 minutes still possible? -16 per 100 possessions.
Odds it gets used a lot (at least relative to others) next season? Pretty good.
Only 4 teams used fewer lineups over 50 minutes last season. Only two more than barely.
Re: Checking in on Celtics lineup usage (and other teams)
I checked the game flows for the playoffs. Game 4 data was missing. In the first 3 games there were 5 spurts were a specific lineup went -5 or worse in a stint. In all 5 that lineup had a good stint previously in that game (short-term memory?) But 3 of these lineups had less than 20 minutes regular season testing (probably well less). The other two were that most used but pretty negative lineup. Not so good to trust key moments to very, very lightly tested lineups... and a quite bad one.
One could spin it more neutral or favorable given competition, but what you can't do is say that they went down relying on well-tested, positive result lineups. Their best 2 playoff lineups (used over 3 minutes per game) were Jerebko lineups. Maybe this was good improvising but it wasn't based on scientific lineup testing as neither was used over 10 minutes in regular season. Good guess or lucky random guess is actually just a guess. But not near enough good or lucky guessing or smart repetition of proven past success.
One could spin it more neutral or favorable given competition, but what you can't do is say that they went down relying on well-tested, positive result lineups. Their best 2 playoff lineups (used over 3 minutes per game) were Jerebko lineups. Maybe this was good improvising but it wasn't based on scientific lineup testing as neither was used over 10 minutes in regular season. Good guess or lucky random guess is actually just a guess. But not near enough good or lucky guessing or smart repetition of proven past success.
Re: Checking in on Celtics lineup usage (and other teams)
Recent ESPN debate on how Celtics can go from good to great.
Another answer not really discussed: Getting to great lineup management. Almost everyone has already decided they already have great coaching. But in the above data I see a ton of room for improvement in lineup management.
Another answer not really discussed: Getting to great lineup management. Almost everyone has already decided they already have great coaching. But in the above data I see a ton of room for improvement in lineup management.