APBRmetrics

https://nbarapm.com/datasets/strong

Among the notable estimates:

Some superstars step it up against the strong but Steph Curry a point lower against the strong compared to weak opponents. Same for Durant.

Kessler 2 pts better against strong.

Lillard 1.6 lower against strong.

Jrue Holiday 60% less good against strong.

VanVleet 40% less.

Holmgren, Caruso, Jalen Williams, Dort better against strong. Hartenstein half as good. SGA 115% less good. (2nd biggest net pt loss in dataset.) Could be defense inspired shift of production / credit. Wallace 10 times as bad?


KAT and Brunson less good. Bridges much better.

Morant and Bane less good. Aldama 3 times as bad.

Avdija less good.

Garland 75% less good.

Randle more than 2 times lower.

Westbrook way worse.


These estimates don't necessarily track with statistical performance in playoffs compared to regular season. Some huge contrasts.

Hachimura twice as bad against strong.

Herro 2.5 times worse.

Naz Reid 5 times worse.

Jalen Green almost 4 times worse.

Splits are more subject to randomness than whole but this is a 7 year sample so length should more than offset that element in most cases. Would help to hear about how strong and weak are defined.

Jalen Smith 2.5 times better.

Cade Cunningham substantially less negative.

Anunoby 40% worse.

Sochan twice as bad. Vassell down some too.

Trae Young more than 4 times worse.

DeRozan more than twice as bad.

J Grant more than twice as bad. T Harris twice as bad.

Conley 40% less good.

Smart 1/3rd as good.

S Sharpe almost twice as bad.

Sabonis loses 90% of value.

So does Siakam.

Gobert loses almost half of value.

Critically important dimension getting some needed attention here and in Mike G threads on RS to PO performance.

USING this list has a ton of apparent value. Applied analysis. Could make or break seasons and jobs.

What teams produced this on their own? Who used / uses this product? Who reads / values this quick review?

More cases to see but these 40 say a lot. Insights into the relative wisdom of many hundreds of millions of contracts in less than 1 hour.

Most of these don't surprise me. A few do because of the degree of change.

That is an interesting link. Sorting by "Strong minus Weak" I see Shai has been 0.6 pts worse than avg vs the Strong.
I don't see a definition, but maybe 'strong' = 'above avg'?

At the better vs strong players end of things, it looks pretty random. Top is Danuel House (+3.0), then Tony Allen, Otto Porter, Ayo Dosunmu, Gary Trent Jr., (w >1000 min.)
At the low end are about 33 players as far as 3.0 from the norm.

Sabonis loses 90% of value.

Implying an avg (top 120?) player has zero value?

Stating that Sabonis was only 0.2 better than the league mean against strong teams and +2 against the weak by this methods estinate, subject to some error (presumably modest or moderate). League mean being obtained fby a relatively small minority.

Raw +/- on / off happens to be +1. right in middle.

0.2 against strong ties him with LaMarcus Aldridge on this.

They come in about 220th in a group that probably has 800-1000 players over the period.

League mean performance is a tough standard to reach. But it is objective and useful, if you want to use it.

Databallr says:

"Strong = opp lineup > +1 Darko" [sum].


It is not that obvious / easy to find an example of a +1 Darko lineup among starting lineups. A point worse than Sixers conventional starters without Embid. A point better than conventional Raptors starters. The Griz starters with Wells and Pippen instead of Morant.

Players not so good against strong, if considered, should be considered more by meh to bad teams and the east, which overlap to a large degree.

Should not be considered or retained by western contenders.

Strong / weak opponents is one of many, many possible splits I have mentioned in past.

A new one might be to split the data into 3 or 5 piles by their impact on the RAPM estimate. Is there a pattern or multiple patterns in the most "helpful" plays and the most harmful? Are there a block of plays that on its face don't look meaningful but RAPM "thinks" they are? Who is "right"? What exactly is happening or not happening on these cloudy plays? Apply factor analysis or other tecniques?

RAPM by conference is another obvious and useful / necessary split.

https://x.com/bballstrategy/status/1912175595246829788

SGA not so hot in 2.25 or 2.5 games of series. An early concern.