We know what didn’t work, but what things DID work?

That lineup was mildly negative on average during the regular season but mildly positive against the Lakers then.

Brooks hasn't used it yet, but he probably will and that be alright by me.

How's this for a lineup to try (might be a bit undersized but could be used when Odom is in and Pau or Bynum is out):

Westbrook (fisher)
Harden (kobe)
Durant (artest)
Green (odom)
Ibaka (bynum or pau)

Thoughts?

his sneakers totally worked...

http://www.kicksonfire.com/2010/04/20/nike-kd-ii-2010-nba-playoffs/

Nate McMillan lost Roy. Said he was up all night thinking about his lineups before game 1. What lineup did he choose to use the most? Miller- Fernandez- Batum- Aldridge- Camby. A lineup that had only been used 38 minutes previously... but happened to have the best statistical rating. So he used it 20 minutes and that lineup won its time by 9 points and they won the game. How did he choose it? By memory and genius or from the stats or a combination? I don't know, but he choose the right one.

He also choose to close the game out with a lineup that he had never used. It worked well too. Good coaching. But I'd rather have some research data on it, all other things held equal.

I didn't read them all but has any body made a g g g g-unit joke yet

kfmsooner, Durant at the 2 and Green at the 3 is something to use carefully, against the right lineups, but it may indeed be more viable now that Durant has learned / committed to defense. At least that is what the small sample data suggests.

Collison may well play better off the bench, as you say, and I agree there are plenty of spots to plug him in later.

More important to monitor the starting lineup's results and reduce its emphasis if it doesn't work as well as others than to specifically change the first lineup used.

But pulling in an upset you don't have much room to give slack or tolerate less than your best.

I'm fine with other folks having their opinion on the topic and their opinion of my perspective but I have mine.

The huge minute starting lineup was good enough to accept during the regular season in general. But as I've noted it actually did poorly against the west and against playoff teams. We'll see how it does in the western playoffs when things are done.

I agree that Collison and Ibaka need to play more. Durant at the 2 and Green at the three was the lineup that got PJ fired a year ago, but I think it is our best lineup.

Also, it seemed when Collison started at the end of the year for Nenad, he didn't play as well. I think he plays better off the bench. Give Nenad the first 6 minutes of the half and give the rest of the minutes to Serge and Nick.

I find the topic interesting. Regardless of what comes of it, other than discussion, I enjoy the subject and don't mind spending some time in it's pursuit.

Some will enjoy it, some won't. That's fine.

... Joe and "I"

Sorry for a few wrong words.

I know I went on about this topic but research about lineup results seems likely to be one of the main ways to improve from where teams are now generally and this team in particular.

The main evidence for "lineup changes" will be the playoff results. The results are just starting to come in. Joe and you both looked at game 1. What is the argument against increasing the use of lineups that worked well in game 1 and the regular season other than leave it to the coach and if he didn't do it he must know better? That it is not enough to me, but if it is for others I can respect that perspective.

I've said wait til after game 3 to make changes and it is almost certainly too late. So maybe the answer is make any changes you are going to make in game 2. If not from the start, then by halftime, if you need it. That is my perspective.

There is a difference between high confidence in the numbers and having the numbers and considering them. Despite how it might sound I don't have that high confidence in most numbers. You are right that the best quants recognize there limitations. But I want the data to consider and nothing substitutes for how a lineup will perform than having it play in real NBA games and the more minutes the better understanding of these lineups and I would focus on the best lineups and gather meaningful data beyond the biggest lineup.

If you use a dozen or dozens of lineups in the playoffs you can either completely trust a coach to select them from memory and experience (lets see, I remember that 2 minutes here, 3 minutes there, etc., etc., etc., I remember everything important that happened on every use of that lineup and every one of the hundreds of lineups I used, so I know exactly which lineup to use right now in this particular moment. Yeah right. I don't have a high confidence in memory and intuitive genius.) or you can use organized summary and study of the data to help him.

The first month they went with a dash of Ollie & Thomas. I can understand and accept that. Presti=Brooks had a reason for this and it was sound. Then they adjusted by deciding they didn't "need" or want to continue doing that.They brought Maynor in and began to ramp up Harden some. Presti=Brooks had a reason for this and it was sound. Then they decided to stand pat, because it was working as well as hoped or better. Presti=Brooks had a reason for this and I can understand it. I still think that after getting to this level then you more on to intense thinking about playoff preparedness and I'd rather have more info on lineups #2-5 or 10 than they got. If you think there wasn't time or this is too critical, I can hear that but I think they could have won as much or more and accomplished this. If we disagree on this, then we disagree. Then the question becomes do you run next season essentially the same way they did this season or do you run it in a way that gets more research data on lineups #2-5 or 10. I know my answer.

"More importantly, you write as if there is no degree of confidence from Brooks’ experience, or film study, or player reaction. You write as if morale and player psychology are immaterial."

If that is how you read what I say and don't say, you have a different view of what I've said and not said than I do. I've put plenty of qualifiers and on the other hands in most of my writing; guess I could put even more, but I am not sure it would prevent this reaction.

I believe in film study. I believe most coaches know a lot about their team but are not at their absolute max efficiency- few organizations get there, but they strive for it and the ones that strive the hardest usually go the furthest. Yes working with real players requires communication and maybe some adjustments from a simple plan. I am willing to give a coach room to vary, but I'd want more rigor in testing lineups in a season than what they do now. This is a highly paid job in a highly competitive industry. Teams that win a lot and go far "players" get orders, make sacrifices, demonstrate things to the higher ups, etc.

One thing we do know from a rather large sample size is that our starting lineup isn't that good. We also know that a Green/Krstic tandem isn't that good.

Maybe the other sample sizes are too small to draw a high-confidence conclusion, but the large sample we do have says that some of what we're doing isn't the best available option.

I'd indeed use the regular season to “get data”. It is a research project- learn what it takes, what you can successfully do, what lineups you can / should use to win against western playoff level teams (mainly). You don't think the best coaches and teams take this perspective at least partially? I do.

"You can’t say it’s a “modest” sample size so you have “modest” degree of confidence." No, the bigger the sample. the confidence increases by more. So a 200, 300, 400 minute sample is much better than the 25, 50, and 100 minute samples. You I'd use the regular season to go bigger, better data samples.

"the quants don’t believe that basketball statistics are anywhere near that level." Either the quants won't admit/ show it in public or they aren't that good. I'd think you'd be surprised how far the best quants take it and the best quants of the future will take it even further. Need to have humility and perspective from other approaches & people. But no, I would not stop short of this level of rigorous lineup research & evaluation. It is absolutely fundamental.

Nice break down Joe. I'm in favor of less Nenad and more Collison or Ibaka.

Speed kills the Lakers. Or so it seems. I would like to see a three guard line up. Perhaps EM RW TS or RW JH TS could be used. That would keep Thabo on the floor to defend Bryant. Perhaps even sneak KD up to the 4 for a few minutes.

I uploaded KD's ankle-breaker on ron artest on youtube: http://www.youtube.com/watch?v=-t0qWgUl5O0

Crow :
And note that I suggested using the regular season as hard as possible to get 100, 200, 300, 400 minute samples beyond the main lineup. If you don’t have more than 1 200 minute sample to make playoff decisions off of, it is because of your lineup choices in the season.

This is not a statistics paper. You don't use the regular season to "get data". You use the regular season to win games, improve as a team, and prepare for the playoffs. Are you really going to tell Nenad Krstic, "You're coming off the bench for the month of February so we can get our lineup samples in order?" That's what you're going to have to do.

Crow :
But under my season long lineup distribution proposal you’d get 10 lineups over 200 minutes. 200 minutes is still a modest sample size and will at best only tell you that it is likely that a lineup is not good or bad (and not for lineups that show in the middle) but it is better than the degree of confidence teams have with almost all lineups now.
That's not true. You can't say it's a "modest" sample size so you have "modest" degree of confidence. That's not how statistical analysis works, with a one-to-one correlation between sample size and statistical significance. It's like saying "we had a cold winter in Oklahoma this year, so that's evidence that global warming is BS."
More importantly, you write as if there is no degree of confidence from Brooks' experience, or film study, or player reaction. You write as if morale and player psychology are immaterial. It's all stats to you -- but even the quants don't believe that basketball statistics are anywhere near that level.

Vince :
The problem with all this analysis is that the sample sizes are so small. There’s nothing in here to override the fact that Brooks has been with this franchise for three years, and knows these players better than anyone. It’s Game 84 tomorrow night — you ride the horse that got you here. I agree with a couple of changes here and there (I, too, did like the big lineup we closed Game 1 with) but these ideas about changing starters and drastically changing minutes don’t make sense to me. Understand how difficult NBA defense is, how based it is on communication and feel and rotations, and then understand that you can’t just throw together 5-man units that are unfamiliar with one another. (The big unit, as unfamiliar as it is to us, still has Collison/Ibaka in the post and Westbrook/Durant/Green outside, all of whom have played together quite a bit.)

holdmymartian :@Crow

I believe, especially with a young group, that Brook’s lineups should be extremely flexible.

I would argue it’s exactly the opposite. Savvy, wily veterans who know their roles can play 25 minutes one night and 3 the next (assuming the communication is there and there is some understood basis for it). The last thing a young guy under the spotlight of the playoffs needs is to be looking over his shoulder every time he commits a foul. There is simply no evidence for the constant calls on this board for lineup changes. This team had a literally unprecedented improvement this season, particularly on the defensive end — where communication, familiarity, and repetition are MOST important. After all that, we’re creating lineups based on 25- and 50-minute samples?

Not just the small samples, but when you watch them on the court they seem to play well and it has good balance with each player. For ex:

Maynor/Harden/Durant/Ibaka/Collison is a great lineup basing it on advanced statistics, but when I watch them play together the stats back them up. The offense is good with this unit based on stats but even without stats you can see it has good balance both offensively and defensively. I don't know why they get so little minutes when they do so well when on the court. 151 minutes total this season and a +86, that's ridiculous. Why not try that lineup more?

Brooks-Presti believe in lineup concentration more than most, but mainly on the very top lineup. I'd just like to see that broaden out to especially emphasize at least the next 3-5 lineups more. Because one lineup is not enough.

But I'd try to know as much I can about 10 as opposed to trying to know and remember and use 50, 100, and really on average 200-300 or more as teams do.

Well to correct myself, they did manage to cross the 200 minute threshold on a second lineup and nearly a third. But under my season long lineup distribution proposal you'd get 10 lineups over 200 minutes. 200 minutes is still a modest sample size and will at best only tell you that it is likely that a lineup is not good or bad (and not for lineups that show in the middle) but it is better than the degree of confidence teams have with almost all lineups now. I'd rather know more about 10 than 3.

Vince, I'd rather a team adjust some after November, some after December, some more after the All-Star break and then probably stand pat unless something blows up or out. And note that I suggested using the regular season as hard as possible to get 100, 200, 300, 400 minute samples beyond the main lineup. If you don't have more than 1 200 minute sample to make playoff decisions off of, it is because of your lineup choices in the season.

It seems like Ibaka gives the team its best energy and they play well with him on the floor but he is saved for the 4th quarter or something. Which works fine unless you don't stay in the game the first 3. Even if it's just for a couple minutes earlier on, get Ibaka some early burn and hope he doesn't pick up dumb fouls. If the Lakers are pulling down rebounds left and right, why not put Ibaka in to see if he can steal a couple? And also hope Harden hits open shots.

The problem with all this analysis is that the sample sizes are so small. There's nothing in here to override the fact that Brooks has been with this franchise for three years, and knows these players better than anyone. It's Game 84 tomorrow night -- you ride the horse that got you here. I agree with a couple of changes here and there (I, too, did like the big lineup we closed Game 1 with) but these ideas about changing starters and drastically changing minutes don't make sense to me. Understand how difficult NBA defense is, how based it is on communication and feel and rotations, and then understand that you can't just throw together 5-man units that are unfamiliar with one another. (The big unit, as unfamiliar as it is to us, still has Collison/Ibaka in the post and Westbrook/Durant/Green outside, all of whom have played together quite a bit.)

holdmymartian :@Crow

I believe, especially with a young group, that Brook’s lineups should be extremely flexible.

I would argue it's exactly the opposite. Savvy, wily veterans who know their roles can play 25 minutes one night and 3 the next (assuming the communication is there and there is some understood basis for it). The last thing a young guy under the spotlight of the playoffs needs is to be looking over his shoulder every time he commits a foul. There is simply no evidence for the constant calls on this board for lineup changes. This team had a literally unprecedented improvement this season, particularly on the defensive end -- where communication, familiarity, and repetition are MOST important. After all that, we're creating lineups based on 25- and 50-minute samples?

While lineup data is useful, play calling on offense and defense also affects the results and can vary between lineups so it is not necessarily all equal other than personnel. You aren't going to get good results consistently unless the lineup can work well together and the play-calling actually achieves it. Bad play-calling can knock down any lineup.

Eric this site allows you to do what you described

http://popcornmachine.net/cgi-bin/gameflow.cgi?date=20100418&game=OKCLAL

Off and on I talk about pairs and triplets or even quads.
Perimeter triplets and interior pairs seem more important, more like functional sub-units to me than the other more random groupings.

These sub-units get more minutes than 5 man lineups and might be quite valuable to evaluate as well. The raw data is not as good as the Adjusted +/- data but I'll use both as I can get it.

But while I fully support having and presenting alternatives and think they often can be strong, it is also important to also spend time asking why they did it the way they did it, trying to find and understand the answer. Was it more than just habit, did it make sense in the moment? You can do it at every lineup change or perhaps it is more realistic in 4 minute blocks or even at the level of quarters. Of course this is individual choice and depends if and how much this fits into your overall way of experiencing and enjoying the game.

Really interesting article. One thing to remember in all of this though is that oftentimes the lineups that are used and the success that they see are dictated by what the other team does. Like when the Lakers had to put Artest back in when we put in Durant. It would be interesting to see a graphic of when each of these nine lineups was used throughout the game and what lineup the Lakers had on the court at that time, along with how the two fared against each other. I'm too lazy to do it though :)

There are some who see lineups just as 5 individuals and on the other extreme some who see 5 man lineups as unique production units. Or you can blend together some of both.

Joe highlighted specific lineups as well as offering general tips on who to emphasize.

Perhaps. Coaching is certainly an art. It can be partly science too. The right way is what works.

@Crow
I believe, especially with a young group, that Brook's lineups should be extremely flexible.

I don't see any major match-up red flags with going with the rotation flow that dreamcatcher suggested. It is going to be challenging whatever you do.

The Lakers under Jackson staffed about 75% of all their playoff minutes last spring with 10 lineups over 20 some games.

Every team and situation is different but I tend to think that most teams could achieve this, if they really know their team well, and might do better with this level of concentration than what most do. I can't "prove" it but I can summarize the data and it suggest it.

Brooks barely used an 11th lineup in game 1. Will watch how that holds up and how the ones he chose do compared to the others used or not used.

Only 2 teams played a lineup an average of 15 minutes per game during the regular season (the Thunder being one, only topped by Memphis) but on the first day of the playoffs 6 of 8 games played their starters 15+ minutes with Chicago using theirs the most- about 22 minutes.

Lineup concentration in the regular season may be tougher to achieve or tougher to sell than I'd like or think wise but in the playoffs it seems most teams get it and do it at least more vigorously for the very top lineup, though they could probably still be more vigorous. (Haven't pulled the data together about lineups 2 thru 5 or 10.)

Heavy starting lineup use is especially true in the 1st and 3rd quarters. Sometimes the 4th. Least common in the 2nd.

Of course it is key that the lineup you give big minutes to actual works. Most especially against the immediate specific match-up or team but you end up having to use what data you have and look at the results for all the minutes you have, the average. you can also dig into and evaluate to some degree the split data against teams or lineups of similar style or quality.

The rotation that dreamcatcher proposes is realistic; it can work minute-wise for everybody. Maybe the 4th lineup is open to other options and maybe Brooks should be able to use a few more but the main idea is go with your best as much as possible.

I've lamented how lightly almost everything but the starting lineup has been used but you can't change that now. Either go with small sample good lineups or put your coaching and management expertise on the line and choose the lineups however you will but be prepared to be judged on the results and see comparisons of what the best lineups according to the stats do vs the others.

Still to be the most realistic we need to look at what the Lakers are doing and how the Thunder lineups match-up. It might change but in game 1 the Lakers basically did this:

Fisher-Bryant-Artest-Gasol-Bynum
Fisher-Bryant-Artest-Odom- Gasol
Farmar-Brown-Artest-Odon-Bynum
Repeat cycle (several times, pretty much)

Close game with
Fisher-Bryant-Artest-Odom- Gasol
(actually the weakest of the 3 main lineups and the other two are strong)

I have assessed the match-ups directly, yet. I'd say that would be the next step.

The first goal in the regular season is to win enough games of course but the second goal (especially in the future now that you are pretty good) should be to learn what you need to know to make the best choices for the playoffs. I think the following minute distribution might be a desirable pattern:

lineup you feel most positive about 600 minutes
next two about 400 minutes each
next two about 300 minutes each
next 5 about 200 minutes each
remaining approximately 25% of total time as you want and / or need to.

No team fully accomplished the amount of time my distribution would devote to the top 10 lineups but the Lakers probably came about as close as anybody else.

I like lineup I.

I got the minutes wrong, add 2 minutes to TS

I think it's pretty clear though that JG and NK is a horrible combination on the front line.

Great article.....

our 3 best lineups are.....
EM, JH, KD, SI, NC
RW, TS, KD, JG, NC
RW, KD, JG, SI, NC

which seems pretty reasonable when you watch them on the court.

maybe we could start the game with.....
RW/TS/KD/JG/NC
couple minutes left in 1st sub in with this lineup
EM/JH/KD/SI/NK
8 minutes left in 2nd sub in this lineup
RW/TS/JG/SI/NC
4 minutes left sub in this lineup
RW/JG/KD/SI/NC

so minutes would go like this for 1st half.....

RW = 18 minutes
EM = 6 minutes
TS = 12 minutes
JH = 6 minutes
KD = 20 minutes
JG = 18 minutes
SI = 16 minutes
NC = 18 minutes
NK = 6 minutes

full game would go like this.....

RW = 36 minutes
EM = 12 minutes
TS = 24 minutes
JH = 12 minutes
KD = 40 minutes
JG = 36 minutes
SI = 32 minutes
NC = 36 minutes
NK = 12 minutes

Maybe if NC has too many minutes replace him for NK for a few minutes, and if JH is playing well reduce JG or TS minutes.

just some thoughts

Russell Westbrook.