2017-18 lineup analysis

Home for all your discussion of basketball statistical analysis.
Crow
Posts: 5491
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2017-18 lineup analysis

Post by Crow » Fri Feb 16, 2018 12:55 am

Report: Celtics to consider lineup changes.

Yeah.

4 of top 7 most used over last 15 games negative, 3 horrendous.

199 lineups used, just 33% cumulatively positive. That is kinda low.


Non-dink lineups plus 19 for this stretch, even with the bad ones.

The 192 dink and super dink lineups? -44 or about -3 per game.

Consider lineup changes? Yeah, at top and bottom. Almost 2/3rds of total time spent in dink lineups, which, surprise (not), cost them points and probably wins.

Crow
Posts: 5491
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2017-18 lineup analysis

Post by Crow » Mon Mar 05, 2018 8:32 am

If you don't like the denomination of minutes per game for the season, then just look at total minutes for the season. The point is low minutes not the terminology. If you miss or dismiss the message over the terminology that is a loss.


Some discussion of Spurs' lineups https://mobile.twitter.com/bballstrateg ... 917824?p=v
Stuff on Magic, probably in mid- January, same kind of checks. A little on the Rockets earlier in season. Maybe someday fans, Coaches, GMs, analysts will do similar or reconsider the issues & patterns discussed or reveal more about what they think & do with lineup data (instead?)

Crow
Posts: 5491
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2017-18 lineup analysis

Post by Crow » Thu Mar 15, 2018 11:00 pm

Toronto Raptors... some good lineup management. Enough?

Top lineup used about 10 minutes per game for season. The right way to look at overall usage. Plus 11.2 per 100 possessions. Could be used more... but top 5 most used lineups are also plus 11.2 on average. 17 minutes per game. Could be more...

371 other lineups used at plus 5.2 per 100 pos. Only 33% of these dink lineups were cumulatively positive. Some greater selectivity and / or lower use of dink lineups could make a difference.

Last playoffs the Raptors concentrated on top 5 most used lineups almost twice as much as they are right now. It helped them barely get by the Bucks. Blown out by Cavs. Lineup tightening could have been more massive and wouldn't have been enough. But this season? It might be the crucial difference. I'd concentrate on best and biggest lineups more than they or almost anyone ever has. Imo, that is the best general, initial strategy out there. Change as information suggests you should, but don't dink up the playoffs as much as everyone still does. Casey is only picking 33% winning lineups. That may not be enough. Last playoffs their 5 most used lineups averaged plus 4.5 per 100 possessions. The rest of lineups, -5.7 / 100. But the dink lineups got twice as much total time.

Don't repeat that Dwayne. Some insider tell him. Do they know? Do they care / agree? Will they see this, follow the suggestion? Probably not. Insiders. Set in ways, not searching enough, not trying new enough. Probably.

Does dink lineup management (both quality & quantity) decide many series including the Finals? I dunno but probably. They get as much or more minutes in playoffs compared to the top few. NOBODY HAS EVER PUBLISHED SUCH A STUDY IN PUBLIC (to my knowledge), or maybe even asked? Have any or many insiders? I kind of doubt it is many. Any? I'd hope so but I dunno. Insider lineup management looks so rough, undisciplined in general from the patterns and results that I wonder about the quality of the analytic advice and / or the utilization of that advice.

Now it may be that coaching moves REQUIRE dink lineup responses a lot. I doubt it is better than using 5-20 lineups but that could be tested. The right person / model could test these alternative strategies. Has anyone? Even if you SHOULD dink, are they choosing the right dink lineups? About 2-1 no on average. Machine learning might improve the selection process, with non-dink and dink lineups. Probably not worse than coaches going on gut and being right 1/3rd of time, plus or minus a little.

Crow
Posts: 5491
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2017-18 lineup analysis

Post by Crow » Mon May 28, 2018 6:27 pm

Celtics vs. Cavs, Stevens uses just one lineup more than 2 minutes per game for series. It was fairly negative. Win rate for 46 other lineups is pretty average. Not good enough. Probably not concentrated enough against Cavs. Some worse teams? Yeah, got by them.

Was there only one lineup he had enough confidence in to play 14 plus minutes? I doubt that. That would be sad after a season of time to test. But only one lineup actually got that meager level of use. Completely weird and unacceptable to me but not to them I guess.

Used about 13 lineups per game. They were so "good" and so "necessary" that on average he used them less than twice in the series. That is a lot of custom match-ups. Usually a losing game and ultimately it was here for them.

Cavs with 3 lineups over 3 minutes per game. Not enough imo but enough against and compared to Celtics. One lineup got 11 minutes per game. Top 3 lineups yield over 370% of the total net points needed in the series. Coaches scrap fight with dink lineups for most of time but the real muscle came from the top lineups. Will teams pay attention, change any? Not seeing much sign they will. It is not just this simple but there is a message here, I think.

Crow
Posts: 5491
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2017-18 lineup analysis

Post by Crow » Mon May 28, 2018 6:54 pm

Top 3 Rockets lineups used over 2 minutes per game in series -28. Warriors top 3, plus 26. Not the full story but a huge part of it. If Paul can't go they won't use those 3 lineups. Might not be as terrible as thought. Those 3 weren't working on average.

My guess is both coaches will use a lot of dink lineups in the decisive game and most of them won't work. The best bigger minute lineups will probably decide the game.

Dink lineups apparently kept Rockets in series but it is usually a losing proposition. Has D' Antoni been brilliant or lucky with dink lineups? Has Kerr been terrible at them or unlucky? Some combination. But again, the bigger minute lineup more likely to provide the final decision than the dink war. We'll find out pretty soon.

Crow
Posts: 5491
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2017-18 lineup analysis

Post by Crow » Tue May 29, 2018 2:30 pm

BRef hasn't posted the lineup data yet but from gameflows it sure looks like Warriors had the better performing top 3 most used lineups. More concentration on both sides. If it made sense to do that, it made sense to do it more earlier too.

Crow
Posts: 5491
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2017-18 lineup analysis

Post by Crow » Tue May 29, 2018 5:30 pm

Rocket lineups over 4 minutes in game 7, -7. Warriors plus 20. Slight win for D'Antoni in the dink wars but who cares. Big minute lineups settled it.

Crow
Posts: 5491
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2017-18 lineup analysis

Post by Crow » Sun Jun 03, 2018 1:31 am

6 of 7 most used Jazz lineups were negative in playoffs, some really terrible. All had Mitchell - Gobert. Just outplayed or were they outcoached too? Lineup concentration evaporated after Rubio went down. No clearcut, confident, effective Plan B. O'Neale not bad in boxscore but a total bomb on team plus minus. Burks won his time but only got half as much, when he played.

Crow
Posts: 5491
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2017-18 lineup analysis

Post by Crow » Tue Jun 05, 2018 5:50 pm

Almost nothing is working for Cavs against Warriors at cumulative lineup results level... except some trios involving a selection from Green, James, Korver, Nance and Clarkson. About 40% of total minutes have such trios. Put all five together? +27 per 100 possessions, though only used 14 minutes. Emphasize trios or all five together more? Maybe. Might be best bets. It doesn't always work but it appears to be working better more often than the 60% of tine without such trios.

2/3rds of Warriors bigger minute lineups are positive. All of them that are not involve at least 2 bench players.

Crow
Posts: 5491
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2017-18 lineup analysis

Post by Crow » Wed Jun 06, 2018 12:40 am

Turns out between game 1 and game 2 Lue cut use of such trios by almost half. Searching for better than that? I guess. But didn't find it.

Crow
Posts: 5491
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2017-18 lineup analysis

Post by Crow » Wed Jun 06, 2018 1:38 am

Raw plus minus isn't a great tool but fwiw Simmons and Saric were weakest on that of the main Sixers rotation. Could be poor play, strong matchups or poor use or a combo. Whatever it is, they'll want to do better there and everywhere but maybe especially there.

Crow
Posts: 5491
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2017-18 lineup analysis

Post by Crow » Mon Jul 09, 2018 9:34 pm

2017-18 Thunder lineup analysis:

414 lineups used. For an average of 3.6 games each. 31% cumulatively positive.

5 used more than 2 minutes per game for season. Intended starting lineup was used a lot and great but derailed by Roberson injury. With substitution in of Brewer or Abrines, just neutral. Ferguson, mildly negative. Huestis, very good. (But haven't heard a word about retaining him.)

If Anthony leaves, Thunder will be either playing lineups used lightly, very lightly or not at all before. Nothing played more than 1 minute per game for season and positive still possible except for one all bench lineup.

Crow
Posts: 5491
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2017-18 lineup analysis

Post by Crow » Thu Aug 02, 2018 12:54 am

All of top 10 most used Nuggets lineups were cumulatively positive last season... but none were used even 4.5 minutes per game for season and the top 10 weren't even used for 40% of minutes. This was ridiculous. Heavy use of dink lineups pulled them down to a non-playoff team. Make that super dink lineups. Lineup used less than 30 seconds per game cost them around 200 points and wiped out almost 2/3rds of the edge the top 10 produced. Unless they radically changed this coaching behavior and lineup trend, they will underperform.

Crow
Posts: 5491
Joined: Thu Apr 14, 2011 11:10 pm

Re: 2017-18 lineup analysis

Post by Crow » Thu Aug 02, 2018 2:39 am

If I was a GM, the Coach and I are going to agree on 10 lineups to try to use 80% of the time. If he refuses to do that, he is gone. If he blatantly disregards that he gets one warning then he is probably gone. If he wants to adjust what 10 are priority, we discuss and agree on it. I would not put up with as much ineffective freelance dink crap as most coaches do and GMs tolerate. I could still decide how much leeway to allow him in picking the 10 but we are going to pick 10 and he is going to play them. Game level lineup usage & timing would be monitored. Any game that does not comply with the 80% is going to be a report or phone call discussion. If the variance for a month is too high... either the results better be really good or see ya.

DSMok1
Posts: 850
Joined: Thu Apr 14, 2011 11:18 pm
Location: Maine
Contact:

Re: 2017-18 lineup analysis

Post by DSMok1 » Thu Aug 02, 2018 1:34 pm

Crow wrote:
Thu Aug 02, 2018 2:39 am
If I was a GM, the Coach and I are going to agree on 10 lineups to try to use 80% of the time. If he refuses to do that, he is gone. If he blatantly disregards that he gets one warning then he is probably gone. If he wants to adjust what 10 are priority, we discuss and agree on it. I would not put up with as much ineffective freelance dink crap as most coaches do and GMs tolerate. I could still decide how much leeway to allow him in picking the 10 but we are going to pick 10 and he is going to play them. Game level lineup usage & timing would be monitored. Any game that does not comply with the 80% is going to be a report or phone call discussion. If the variance for a month is too high... either the results better be really good or see ya.
Sample sizes are so small it's hard to say that lineup X or lineup Y are what should be played. I think it is important for the GMs to allow the coaches the latitude to do their job without meddling--which I think they typically do. There is a lot of value in respecting each person's skillset. There are so many nuances to lineup usage beyond straight analytical things.
Developer of Box Plus/Minus
APBRmetrics Forum Administrator
GodismyJudgeOK.com/DStats/
Twitter.com/DSMok1

Post Reply