APBRmetrics

Crow wrote: ↑Tue Oct 29, 2024 10:20 pm It is super early but this is 1 way to run the Rockets rotation:

VV JG BR JS AS 20 min
VV AT BR E A 4
AT JG Tate E A 8
VV RS W E JS 4
VV JG W Tate AS 4
RS AT W E AS 8

I accidentally wiped out some of the accompanying detail of this post, trying to quote it a minute ago.

I did several later, probably better
revisions but what can be said of this original set of early recommendations?


VV JG BR JS AS 20 min

Used about 7 minutes per game at about +5.2pts / 100p. About 1/3rd my recommended time, based on the small sample available to date then.

Good recommendation.

VV AT BR E A 4

+48... in 41 minutes. Under my recommendation it would be around 240 minutes. Probably good recommendation.

AT JG Tate E A 8

+43 in 4 minutes. Might be good recommendation. No meaningful or even semi-meaningful testing.

VV RS W E JS 4

Zero minutes test. Can't say, except I would have tested.

VV JG W Tate AS 4

-23 for 8 minutes. Way not enough to judge

RS AT W E AS 8

+25 in 3 minutes.

That's how much followup I can do on this one. Looks good for me, as far as testing shows.

Udoka went with 441 lineups with 6 of 10 most used negative. Go with his output based on instincts and whatever else he actually used or mine? Disregard the tiny samples (his doing, not mine), the recommendation to use the starting 3 times as much probably would have been very helpful on its own with the others promising but pretty much unknown due to the lineup chaos chosen.

Udoka picks 46% lineup winners overall.

4th seed, given 1% chance at title by BRef.

Continue as is or change anything? Anything at all, especially with lineup management? I would.

. VanVleet - D. Brooks - J. Green - A. Sengun - T. Eason

has become the most used lineup in last 5 games and is +34/100p in 33 minutes.

Eason in 3 of the next 6 most used. For season was in just 1 of 7 most used but 4 of 10 most used. A change but still fairly modest. Will watch how sustained/ how good.

I have thought strongly of Eason since before draft and since but guess I got so used to him being deployed on bench to elevate him more in my previous lineup set. Lineup sets can be easily redone, with motivation.

Eason in 3 of 4 most used in last 10 games, with more minutes for the other two. Udoka experimenting. 2 are strong early, 1 weak. D. Brooks - J. Green - A. Sengun - T. Eason with Van Vleet or Thompson worked great. Without either of those 2, relying on Green and adding J Smith, it really did not.

How many minutes will these lineups get before end of season? Some prior testing, mostly with Thompson. It might get to 200 minutes. The other 2 are currently well short of 100 minutes. Udoka has rolled out 162 Eason lineups in mix of 468 total. Could / should imo have been far more concentrated on 20 or 10 or 5.

VanVleet - D. Brooks - J. Green - A. Sengun - T. Eason used about as much in last 5 games as all of the earlier season. So it is a start at concentration. A very late start.

The main quad could be said to have tested to a minimally adequate standard. But it could and could have been more robust.

13 quads tested 200+ minutes. 2 of the 5 quads contained within this 5 man lineup tested to that level. Plenty of contained trios tested decently but only one with Eason in the 20 most used.

4 others just below that level and 17 total in 50 most used, all but 2 positive. The negatives had Smith, so were not contained in this lineup. Expectations for this lineup based on sublineup data could reasonably be high.

Neutral and mildly positive in last 2 big minute uses. Does that reduce the use? We'll see.

Amen Thompson back... so let's go back to starting him in a -8 / 100p on average lineup... Goes -14.5 pts / 100p tonight.

Starters with Eason down to 2 trivial minutes.

Should starters with Thompson be a major lineup in playoffs or the biggest of them all?

Lineup data, in modest test, suggests probably not at best.

Should starters with Eason? Maybe. But with inadequate test, it would a risk. A bigger risk than it needed to be and may continue to be because of inadequate testing.

If it is almost April and you don't really know on one potential main lineup and are going against the data to date on the other more favored, what have you learned / gained?

Not much at lineup level, near the top or anywhere under.

Pair data is good. But all 9 of the best in 20 most used are 8th or lower and the 6 worst in 20 most used are all 1st to 9th. Yes this is raw data but it is really optimized? Does not seem that optimized on the surface.

Either you are satisified with the results and process or you want to try for better and better before the playoffs for possibly knowledge that can yield an extremely critical boost. Even 1-2 pts per game from adjusted lineup management could have huge impact.

Starters with Thompson -30pts / 100p in 14 minutes tonight.

With Eason, a few seconds or a FT lineup change.

Other things worked and didn't but the starters with Thompson negative overrode the sum of everything else almost twofold.

No Thompson yesterday. Eason back with starters. +13 / 100p in 11 minutes.

Probably back to Thompson when he returns.

Different preferences. Probably different approaches.


Eason positive with entire main rotation of players on average for season. Last night it was with starting unit and probably some of the rest of the time with VanVleet. But apparently really not with the handful of non VanVleet time.

It can vary game to game. But starters with Thompson tonight lost their time mildly.

Starters with Eason got 1 minute. Lots of things tried but that isn't getting consistent looks.

Will the Rockets go into playoffs using a starting lineup with Thompson that was -10pts / 100p?

Go back to Smith after moving him to the bench?

Use one of two Eason lineups tested only 100 and 150 minutes each?

Is this how you thought the regular season was best run?

Rockets starters with Thompson, -7pts / 100p tonight, in line with regular season results. Out of line with my previous suggestions.


2nd most used lineup had plenty of earlier testing and worked.

3rd most used lineup had 17 minutes of regular season testing. It worked.

4th most used lineup had 100+ minutes of testing, looked good then and worked.

5th most used lineup had just 29 minutes of testing, less slightly negative but worked today.

+8 with 5 biggest lineups. -18 with the other 10.


Smith - Eason, close to least good Eason main pair. 15 bad minutes tonight. Eason with Adams and Van Vleet, modestly better them median in regular season, best Eason pairs tonight.

Some things look good, others not. Do less wise and get weak results and it can add up and overwhelming the good.

10 smaller lineups (generally or all sparsely tested) pulling -18 is the general tale.

Starters with Thompson had some success in recent games but 12 minutes at -53pts / 100p massively lost this game and pulled the average out of positive territory. What happens next game?

Of the next 5 most used Thompson lineups for series to date, 3 were very large negatives and 2 were neutral. Pretty large to very large negative pairs with everyone but Adams. All were non-positive before tonight but most only mildly. -17 tonight, clear worst. Not an issue in boxscore tonight. Shooting / scoring efficiency was awful on average in previous games.

7 of 10 most used pair negative. 5 terrible.

9 of 10 most used trios negative. VanVleet Brooks Sengun was the positive and most used. The next 9 were almost all bad to terrible.

Thompson and / or Green in ALL these bad to terrible pairs and trios. VanVleet and Brooks in about half. Even Sengun gets dragged down in some.

Overall playoff +/- on court, it is Adams +14/100p, Segun and Eason near +2 but with strong on / off. Smith meh. VanVleet, Brooks, Green increasing negative on both measures, Thompson by far the worst of the main rotation.

Series not over yet; but I'd say off-season would benefit from reconsideration of almost everything.

Sengun Adams was +31 / 100p in regular season in about 150 minutes. Still +25 in 12 minutes per game in playoffs. +38 in 23 minutes last night. Too many other things were terrible to horrendous to save that game but they should try that pair for that long again.

Rely on it in regular season? An open question. Would have to re-sign him first.
Which playoff contenders bid on him?

Starters with Thompson back into the positive overall. But another game comes.

Top 3 lineups are positive but then next 7 are 5 negative (4 badly), 1 neutral, 1 positive.

Most used pairs look better than expected for a team behind on wins. Only negative in 10 most used is VanVleet - Green followed by 2 Thompson pairs with VanVleet and Brooks at barely positive. Rockets perimeter is the weak spot.

Brooks - Sengun is the clear leader at almost twice the positive of 2nd best in group.

Starters without Jalen Green and Amen Thompson were +46... in 18 minutes. Mildly negative in regular season... in only 78 minutes. Use it more or way more? It would be a guess because of inadequate testing... but I'd try it. It is the best with and without combination of the starters over 5 minutes. Not used at all last game and just barely in last 3. Hit upon and forgotten.

Each combination of Sengun with 2 of Thompson Smith and Eason in playoffs was positive overall as was each combination with just one. With all 3 only got 8 minutes and flopped but 8 minutes is too short to judge.

Without Sengun was mixed but some combos worked including with all 3 of the wings.

There is much more to assess.



This is jumping around but Jalen Green was in all 6 of the 20 most used quads in regular season that performed at +1 or less. Brooks, VanVleet, Sengun and Thompson in 4. The combos mattered then. Will try to match up the playoff data later. There were 110 quads used. The 50 most used were 3 to 2 positive. The smallest 60 were 2 to 1 negative.


The 5 most used lineups were +47. Everything else (46 lineups) were a net -47. Concentration matters. Those 5 lineups played an average of 23 minutes per game. I was able to verify that they could have played 39 minutes per game and stayed within the minutes given to each player with one minor variance. The last 9 minutes would still be a mixed bag without VanVleet but with some of the others as it was. Could have played the 5 most used lineups up to 70% more... but did not.

Udoka did cut down to just 9 lineups in last game (so it can be done) but it was much more on average in previous games and the 9 was 3 positive, 6 negative this time. Maybe concencentration would have made a decisive difference in an earlier game? 17 lineups used in close loss on April 28. VanVleet's time split 12 ways instead of just 5. 6 most used, all positive. 6 smallest, all negative.