Things going swell and they have a super lineup that they use a lot. Yay! But not a single other lineup with more than 75 minutes testing, or the equivalent of 2.5 minutes per game for season. Will they know from data what supplemental lineups to use and not in playoffs? Not so much really. Have to rely on coaching feel again (are they confident that is an adequate strength?) What if super #1 lineup is not available? Pretty much guesswork. Sure was too bad all that testing tine got spread out of 213 mostly dink lineups so far. Sigh. I guess this is not a blunder if you don't think lineups per se really matter. Seems like most NBA don't, right or wrong.
Griz almost exact same story, so I won't make a separate thread. Bulls and Warriors just slightly better at testing secondary lineups. Houston? They have below avg. minutes for big lineups compared to those I've checked. Clippers, Doc has just 7 lineups over 20 minutes. 2 getting adequate burn. "Limited" himself to only 141 dink lineups. Such discipline.
Pop? He is up to his usual fun with 11 over that imposing 20 minute threshold and 265 below. Doesn't matter to him. Would it help the guys trying to beat him? I'd think so, but can't prove it yet.
This isn't adequate analysis but first idea / look: the teams who beat the Spurs in playoffs in the 6 years between last titles averaged a bit over 6 lineups used over 100 minutes in regular season and never less than five (only once). For the title years, the opponents had 5 and 2 for an average of 3.5 or 40% less. Small data set that could be mocked even after the disclaimer but all the Spurs beaters had equal or more "big minute" lineup testing, that might have helped some. Data aside, it doesn't sound unreasonable to me that to beat the Spurs it helps to have major lineups with a bit more experience together rather than less.
Checking in on lineups for many other teams
Re: Checking in on lineups for many other teams
Cavs have several Tristan Thompson lineups with good performance, far better than with Varejao. Could work out to increase time for these.
Re: Checking in on lineups for many other teams
I am told Denver is a mess. Lawson - Affalo - Galinari - Faried - Mosgov used 7 minutes per game but in only 1/3rd of Galinari's games. Small sample size but did they have better? This lineup was plus 7 per 100. Too much playing with other stuff? 276 dink lineups under 20 minutes use. Substitute in Chanfler or Hickson for Faried (even better in shorter minutes) but the other 4 are a good group. Maybe not as simple as just this sounds but may not be as hard as they are making it either. Nugget watchers in the house?
Re: Checking in on lineups for many other teams
Is there no one else that thinks 7-12 lineups over 20 minutes (most not by that much) and 200 plus dink lineups under 20 at this point is absolutely ludicrous? Or just no one that thinks that coaches are ever going to change?
Re: Checking in on lineups for many other teams
What's the alternative? Play your best lineups more? How do you know which are best without trying a bunch of them?
You need a lot of minutes to gain statistical significance. You also need a lot of minutes for 5 players to figure out what they can do. There are only so many minutes in a game and in a season.
Complicating this is the fact that an opposing coach makes a change whenever your lineup is going well. Do you then play into his hands then, or counter with a change of your own?
You need a lot of minutes to gain statistical significance. You also need a lot of minutes for 5 players to figure out what they can do. There are only so many minutes in a game and in a season.
Complicating this is the fact that an opposing coach makes a change whenever your lineup is going well. Do you then play into his hands then, or counter with a change of your own?
Re: Checking in on lineups for many other teams
There is almost 4,000 minutes in a season. In the first thousand minutes you can try a couple lineups a 100 plus minutes and a dozen or more for 20 plus. After that you can pick one to turn into a 1000 plus minute lineup and maybe 5 to take over 200 minutes. In playoffs, you need a set of 5-10 lineups to ride unless something unusual happens or stops working. You don't need 500-600 dink lineups. 70-80% of the time probably can be handled by 20 lineups. One or no lineup over 500 minutes and just a couple over 200 minutes is the norm and that has very little testing value for the playoffs, which is what it is all about to me. What I suggest above still isn't a lot of super solid information but it is better. I'd say a lot better. A possible information edge for the playoffs, when the random dink lineups are a bit rarer and probably should be even rarer still.
Re: Checking in on lineups for many other teams
7-12 lineups over 20 minutes at this point, means 7-12 lineups used over 30 seconds per game or over 3 minutes in 20% of games. If I was making a fuss about not many over 100 minutes that would be a bit different but they have so few over such a low, easy to reach standard. If they by eye knew good lineups from bad, you'd expect more over 20 minutes that actual, I think. If play calling calls for specific skills in specific spots and chemistry is important and something that coaches understand more than others where is the evidence of them optimizing this much? Are there really 600 situations in basketball that are optimized by 600 different lineups in a season?
Re: Checking in on lineups for many other teams
One could argue that most teams have done in 1500-1600 minutes what I suggested in first 1000, but even if you let them off hook for part a, most heavily dink away the rest of season instead of increasing concentration of better or at least better guess lineups much. Injuries and trade matter, on some teams more than others, but the gap between the focus that is usually possible and the actual distribution is pretty huge.
Re: Checking in on lineups for many other teams
I think the coach can figure out what lineups are working much faster than statistical significance can be achieved for end of possession outcomes. He can tell if the plays are being run well, players are meshing, etc. Intermediate factors that produce scoring are going to stabilize much more quickly than the points themselves.
Re: Checking in on lineups for many other teams
It is a topic that should have more research on it than my short posts or your I think, I think with no disrespect meant.
On another brief check the ratio of positive lineups used over 50 minutes vs negative ones last season was 55-45. I know the force that pushes it to near 50% but the lineup choices of those intuitive coaches is not making much head way against it. I would think their most common choices would do better than their smaller guesses and random moves but it appears not much.
On average teams end the season with just 12 lineups over 50 minutes for the season or pro-rated to a minimum of 40 SECONDS a game. With 90-95% of lineups not reaching this threshold. I'll probably calculate the percent of total time later when on a computer. But if someone wants to help the research / discussion, that would be nice.
On another brief check the ratio of positive lineups used over 50 minutes vs negative ones last season was 55-45. I know the force that pushes it to near 50% but the lineup choices of those intuitive coaches is not making much head way against it. I would think their most common choices would do better than their smaller guesses and random moves but it appears not much.
On average teams end the season with just 12 lineups over 50 minutes for the season or pro-rated to a minimum of 40 SECONDS a game. With 90-95% of lineups not reaching this threshold. I'll probably calculate the percent of total time later when on a computer. But if someone wants to help the research / discussion, that would be nice.
Re: Checking in on lineups for many other teams
I guess the postulate that coaches need to select a lineup really specific to the moment, opponent and recent performance should be given fair consideration. I've assumed and gathered evidence that leans in other direction but perhaps there could be evidence for it.
If someone was great at handling big data and modeling, one could test the probability of stint winning for every lineup / coaching decision for an entire season league-wide or ten, based on one or more methods of predicting win likelihood. And compare lineups used over 50 minutes vs. less and lineups that analytics would have recommended by used over 50 minutes vs. not or lower basis to recommend.
If someone was great at handling big data and modeling, one could test the probability of stint winning for every lineup / coaching decision for an entire season league-wide or ten, based on one or more methods of predicting win likelihood. And compare lineups used over 50 minutes vs. less and lineups that analytics would have recommended by used over 50 minutes vs. not or lower basis to recommend.