APBRmetrics

Game 41, back to the original Celtics starting lineup. Different big minute bench unit but again negative, wiping most of starting unit edge. Would really help if the coach "knew" / could be guided to or could stumble upon a reliably better one. Back to 24 lineups used, so the 14 in previous game may be aberration rather than new model. Back to introducing 7 never before used lineups. Best performing by far lineup amongst 7 most used for season again goes unused, as it has in 60% of games.

Celtics Game 42, the original starting lineup sucked badly, badly enough to cut their performance positive for the season in half down to a very modest plus 2 per 100 possessions. Big minute bench unit? A bit above neutral but not near enough this time to fill the hole created by the recently restored starters.

That best performing big minute lineup I bemoaned the lack of use of in my last post? Yay, it got used last night. And did great in about a 5 minute stint, plus 11. But then got used just a few seconds the rest of game. Another lineup with a plus 6 stint? Never used again. Another lineup with a plus 5 stint? Never used again. 21 lineups used. Too many, given the strong performers and the lack of adequate encores. If you are going to go tactical vs strategic on lineup usage, maybe give what is working more time?? Now perhaps the opposing lineups changes justify changing from or not going back a winner against a different kind of opposing lineup. Maybe. That would be a new level of analysis I haven't attempted yet.

67% of the lineups with non-positive results. This part was normal. And yet coaches appear to have high confidence in their tactical lineup calling.

Celtics game 43. Playing a good team. Traditional starters have a lousy record against good teams, as reported previously. What does Stevens do? Rolls out the traditional starters. What happens? -14 for them.

More of the hard to accept from last night:

Lineup goes plus 5 in a few minutes in first quarter. Changed slightly at break and never used again. Another lineup goes plus 10... and is never used again. Still another plus 5 lineup, same thing. I am not leaving other very good results out. 3 best producing stints, no repeats. Starters horrendous to open game, repeated for only slightly less horrendous results in 3rd quarter.

Tanking? Unlikely. But looks unwise / bad to me.


Starting lineup now slightly negative for season. Picked it, rode it, went away. Went back. Now what? Stick with it or flitter around again? Great lineup with Sullinger and Olynyk back to unused again.

Big minute bench unit concept used for a few games? Not used last night. Substitute in 6 smaller ones.

3 most used lineups for season. Meh on all three. Slightly negative for season cumulatively. Don't seem to be able to find a good big minute lineup. I guess being plus 20 per 100 possessions isn't enticing enough to try Thomas, Bradley, Crowder, Sullinger and Olynyk more. Why? I have no idea. Even if you refuse to use that one (with a reason or not), there are several other top12 lineups with great results in light testing that could be tested more, consistently.

Team relying on the occasional surprise pot of gold payout at the dink lineup machine.

David Blatt's lineup data:

40% of lineups were positive, or only 60% non-positive. Probably near very top in league in that.
Only 298 lineups used. Probably below average. Also good to me.
Best five lineups were plus 173. Rest were plus 59. That strong double positive is pretty rare.
But within 5 most used it was 2 very good, 2 meh, 1 bad. That was kinda weak.

Actual wins versus expected was a nice enough plus 2.
But wins against top 10 and top 16 slipped from 2nd best on each last season to 5th best this season.
6-5 against top 10 teams may figure a good deal in the decision.

Breaking- Celtics lineup I've been touting with Sullinger and Olynyk gets the start tonight.

Actually got 1 previous start. Went plus 18 on Dec. 23. Then not used as starting lineup again for a month.

A. Johnson - Sullinger lineups produce just 12% of the edge of Olynyk-Sullinger lineups but to date have been given almost 4 times the run. Olynyk-Johnson lineups produce about 350% of the edge of Johnson- Sullinger but have given only one-third the run. The choices defy the numbers in many ways. I'd rather rely on these small sample numbers for the Olynyk pairings for guidance than whatever mix of coaching instinct and analysis that produced these massively different results for Olynyk and non-Olynyk lineups. Of course you still monitor and adjust if necessary.

But yay for giving your by far best performing lineup with over 35 minutes of use its second start in game 44. How is that for speed of analysis and adjustment? Will it get start 3 in game 45? Will it stick? Or does he go back to his starting lineup with 27 tests and a slightly negative cumulative result? Moved away and gone back to it at least 3 times previously, including in recent days. Using something different against a good team tonight got a good result, in contrast to the previous game lost to a good team while using the standard lineup.

I now see Amir Johnson was out tonight. So this change is probably borne of necessity rather than choice.

New starters with an ok stint, a bad one then a great one to win the game. Has to prove itself, but to do that it will need more chances.

What should the Clippers do at this point? Only option would seem to be staggering the starters aside from the beginning and end of games but even that might not be enough.

Crow, I have a quick question about your methodology. Often you mention these dink lineups and how you dislike their use. Have you ever run your numbers to discount lineups in garbage time? I agree that Stevens uses too many lineups, but looking at the worst lineups used, they are mostly 1-2 game lineups and around 60% of them involve at least one of the rookies/deep bench players.

I guess being plus 20 per 100 possessions isn't enticing enough to try Thomas, Bradley, Crowder, Sullinger and Olynyk more. Why? I have no idea.

Do you have any data on the opponent's lineups when this lineup is on the floor? It would be interesting to see if this is mostly against backups or starters. Conventional wisdom would say that this is the type of lineup that would be abused by a good 1-4/1-5 PnR, but the data could say differently.

I agree though, the lack of Olynyk lineups is baffling, he needs to be playing more minutes.

Rd11490, I haven't separated "garbage time" lineups out. I probably won't unless someone is paying me for the analysis and giving me their definition of garbage time.

I could find the opponent lineups for that one great lineup. I might get around to that, because it a good question and not that time consuming.

fpliii, you are right about wisdom of Clipoers staggering starters. The biggest minute all bench lineups are performing badly, should be avoided and can be probably.

Apart from taking GM duties from Doc and trading or sitting most of their wings, they should increase the time of Mbah a Moute and Wes Johnson. A Rivers and Stephenson should only play when there is no realistic alternative, the matchup is favorable or they are hot. Pierce and Crawford only slightly less restricted.

Stevens went back to his traditional starting lineup. A great game followed by a slightly negative one.

More on Thomas-Bradley-Crowder-Sullinger-Olynyk lineup data

First use in game 2. -1 in 1.7 minutes. Next in game 5 at home against Wizards. Plus 8 in 3.6 minutes. 1st quarter, 1st Celtics substitute after Wall went to bench. Fairly small Wiz lineup. Good to see in game 6 but -4 in 1.4 minutes. 5 games later plus 14 in 3.1 minutes against Mavs in early first right after Zaza sat down. After 15 games it had 5 uses, was 8th most used lineup and was 2nd best performer of those. Another Olynyk lineup had been used slightly more.

After 25 games it was 10 uses and not as special. 7th most used. In game 29 it got 22 minutes against mostly starters in 1st, 3rd and 4th quarters. Hornets kinda small. Johnson out. After this breakout game, it got 0 minutes (just a brief appearance on books as transition to something else during free throws or a time out). Bad game against Lakers on Dec.30 but scattered minutes hard to see anything from. Good game January 18 against the Mavs again. Then the second start against Bulls January when Johnson was again out. "Neccessity". Plus 4 in 15 minutes including a big end of 4th, against a kinda small Bulls lineup. Good games seem to be mostly against kinda small lineups. After 35 games, had been successful at different times of game, 2nd most used lineup but at less that 2 minutes per game across all games. 2nd best performer of top 7 most used.

After 46, still 2nd most used but best performing amongst top 7. Need to dig deeper on what is working / not at factor level. But a good case to use / test more. Only used in about 38% of games. If you don't like the low degree of testing, all but one lineup was tested less, most much less. A consequence of using almost 400 dink and super-dink lineups. Success in game 42 and 44 was first consecutive strong games but no use in game 45 or 46. I guess that and being 2nd best big minute lineup fails to be enough to impress to get more regular or greater use. Would a 3rd or 4th consecutive result help? Maybe. And yet mildly positive starting lineup keeps getting called far more despite its ups and downs. The only lineup with a decent test. Not a miserable choice or result but the very mild positive numbers on average don't provide support for choosing it and coming back to it over other choices. Seems like settling.

More analysis of this good Sullinger - Olynyk lineup could be done but the greater need is for more testing. Just 10 uses over 3 minutes and only 5 over 5 minutes. You probably will use it in playoffs and unless you test it more and analysis it more, it would be hard to say much more than appears good overall and perhaps especially at times against smallish lineups (at forward posts). Only positive at game level in 7 or 19 uses, so not more consistently positive than the norm but much much better on average per minute. That is the main thing we have to go by. It is not much but it probably better than anything else used including coaching "intuition". I don't have enough basis to guess what kind of lineups it is not good against given only 2 games with more than 4 minutes use and -3 or worse result. Need more testing. I'd give 150-200 minutes use over next 20 games to see if it should be kept at an elevated level or not. If kept at current rate it will get about 35 minutes of random use.
Unless something changes, by playoff time they will be able to say the starting lineup is meh ( or better or worse) and this 2nd lineup is good in too brief testing and not much else at 5 man lineup level.

If Sullinger is not re-signed, there isn't anything with depth that they learned from this entire season at the 5 man level that they can exactly carry forward. Perhaps headed for another season of not so disciplined and strategic lineup usage. Is this the way to identify / build a bright future? IMO, no. Not even close to what is possible and appropriate in terms of gathering data on key lineups. And they have spent far more time on Sullinger-Johnson lineups than Sullinger-Olynyk and Johnson-Olynyk lineups even though Sullinger-Johnson are by far the weaker performing pair. And Johnson may not return either. The lineup management looks non-strategic in several ways, average at best and disappointing to me.

In 4 games Ty Lue has used 61 lineups. That might be a bit less than average. 67% non-positive or close to average. Top 4 most used are all good to great; but 6 of the next 7 are awful. Will check back in future.