APBRmetrics

By RAPTOR, last season's championship Warriors fully met all four criteria. In 2021, Bucks barely missed on C1 and C2 but made C3 and C4. In 2020, Lakers made first 3, barely missing on C4. Raptors did the same in 2019. Warriors did the same in 2015, 2017 and 2018. Cavs hit C2- C4 in 2016. Modest miss on original C1 but would make it on revised lower standard. Spurs hit all 4 in 2014.

Appears hitting multiple criteria wins titles without exception over 9 years. Guess I should look for any cases where that is not enough to be THE ultimate winner.

2022 Celtics hit C2 - C4. Modest miss on original C1 but would make it on revised lower standard. 2022 Heat, modest misses on 1 and 2 but hit 3 and 4. 2021 Suns missed on 1 but hit on C2-C4. Multiple criteria hits are not necessarily enough.

Using regular season data. Will check playoff data. 2022 Warriors were monstrous on all 4. 2021 Bucks missed on C1 but got the other 3.

It is clear hitting multiple criteria is the norm for title winners.

So... looking good for 2023 Celtics and less likely for other 3 by regular season data. By playoff to date data, Celtics hit 4, Nuggets 3, Heat 2, Lakers 1.

Playoff Sixers barely missed on C1 but hit on other 3. The criteria are not too enough. Maybe need to be tougher to separate "the one". 8 of last 9 hit at least 3. Only 2021 Bucks had less in regular season but stepped up a notch in playoffs.

Will any national media notice and cite this analysis? The latter seems unlikely. Will any team staff read and think about this? I dunno.

Aren't there a lot of recent examples of teams trying to do exactly this? Nab two of the best players and a strong 3rd and ... miss the playoffs?

Of course lots try and some achieve a bit but still miss going deep, as the data I provided shows.

I put more specificity on the criteria / needs.

I wonder if you can build a very very simple model just based on the advanced all-in-one stats of the best, second best, third best player to predict playoff performance

E.g. Take the top 3 regular season BPM, RAPTOR, EPM, DARKO to predict playoff win%. See if you can do it better than standard methods e.g. plus minus + pythagorean

Simple model or more detailed could be interesting / useful. Perhaps more interesting / useful than some other stuff being pursued in media or maybe inside.

From team win projection thread:

v-zero wrote: ↑Mon Mar 27, 2023 1:41 pm My predictions are being seriously flattered by the MAE here, but to add the to prediction averaging chat: whilst I did not use any minutes projections at all for this, merely flattened average statistics at several roster sizes, I did average the predictions between those roster sizes. Some teams showed up as having a great 5 man roster, but lacked depth and as such were dragged down. Some were vice versa.

NBA teams pay much more attention to playcalls than lineups. Would be useful to check play efficiency by major lineups or lineup clusters. Does bballindex.com or Synergy or other source have this data by playtype? How consistent is play efficiency across lineups? If it is not very consistent then lineups matter.

Crow wrote: ↑Wed May 17, 2023 3:52 pm NBA teams pay much more attention to playcalls than lineups. Would be useful to check play efficiency by major lineups or lineup clusters. Does bballindex.com or Synergy or other source have this data by playtype? How consistent is play efficiency across lineups? If it is not very consistent then lineups matter.

Hmm, I agree that play call efficiency would vary greatly depending on lineup. However, without a lot of manual labor we wouldn't be able to go beyond the off-screen/handoff/P&R ball handler etc coarse grained play type that most data providers have. Also I don't know if there's enough data in a season once you split by lineup and play type. Maybe you could characterize lineups e.g. 5 out, 4 out, 3 out etc using 3PT% to deal with data insufficiency problems.

The way Coaches / teams run lineups, only a few 5 mans have even modest practical significance on their own. 4 mans or clusters of all lineups with 4 or more of 5 or 6 guys will have better sample sizes and may be useful, alongside with or without you analysis from that core group.

Lineup data and subsequently lineup management could be much better but won't be without a revolution in coach think or diminishment in coach power in favor of analytic input.

Unwilling to select or impose a radical lineup concentration model immediately on NBA team? Use G league to experiment. Coach / analytic support get 20-30 lineups for 90% of total minutes for season. That's it, no more; so plan accordingly. Document every part of the work process, thinking / reacting and evolving results. If you make it work, use to argue moving up a level. If you don't make it work right away, change more (thoughts / habits / people) til it does or til you give up any hope / belief in the approach (assuming you start out with some). Develop a real lineup testing design & time sequence within the constraints.

Nuggets succeeding with stars still on 2nd contracts and a Coach known for tactics and toughness.

Heat succeeding with some 3rd contracts and some bad contracts, in part because of more than usual effective low contract role players. Coach known for tactics and toughness.

Celtics had the 2nd contract stars but not the premier coaching. Lakers the 3rd plus contract stars but not quite enough cheap / effective role players. Coach maybe not effective enough or tough enough.

Nuggets and Heat both with just 7 -1 or better on BPM in regular season.

Heat stretched to 11 in playoffs. Nuggets stayed at 7.

Depth might be key factor in finals.

Lakers should move on from Russell, Beasley and probably Bamba. None of them had a strong scoring efficiency playoffs, this time or ever.

Lots of others are marginal or worse.

Not sure that Beasley, Vanderbilt, and D-Lo had more positive impact in 3 rounds, than Westbrook had in 1 round.

From almost a year ago:

Crow wrote: ↑Sun May 21, 2023 5:06 pm The way Coaches / teams run lineups, only a few 5 mans have even modest practical significance on their own. 4 mans or clusters of all lineups with 4 or more of 5 or 6 guys will have better sample sizes and may be useful, alongside with or without you analysis from that core group.

Lineup data and subsequently lineup management could be much better but won't be without a revolution in coach think or diminishment in coach power in favor of analytic input.

Unwilling to select or impose a radical lineup concentration model immediately on NBA team? Use G league to experiment. Coach / analytic support get 20-30 lineups for 90% of total minutes for season. That's it, no more; so plan accordingly. Document every part of the work process, thinking / reacting and evolving results. If you make it work, use to argue moving up a level. If you don't make it work right away, change more (thoughts / habits / people) til it does or til you give up any hope / belief in the approach (assuming you start out with some). Develop a real lineup testing design & time sequence within the constraints.

Does any G league team try anything close to this? I doubt it. But has it even been seriously considered / discussed? Why the heck not, besides over-deference to Coaches?