All-Star selections 2018

[Mike G]

How is Andre Drummond not one of the top 12 players in the East?
He projects to net 10.8 Win Shares and 5.8 VORP by season's end. The only other players on course to hit 10.5 and 5.5 are LeBron, Giannis, Harden, and maybe Westbrook.
In the previous 10 seasons, it's been done 51 times -- 5 per year.
Here's the whole accounting of players who would go on to reach 10.5 WS and 5.5 vorp while being left off the AllStar teams:

Code: Select all

year  not star  tm     WS   VORP  pr
2014   Lowry    Tor   11.7   5.7   0
2013   Curry    GSW   11.2   5.6   0
2013   M Gasol  Mem   11.5   5.7   1
2008  Ginobili  SAS   11.1   5.9   1
2007   Brand    LAC   11.5   5.5   2
2002   B Barry  Sea   12.1   6.0   0
2002  B Wallace Det   11.6   5.9   0
2001   Francis  Hou   12.2   7.0*  0
1997  Blaylock  Atl   12.5   7.1   1
1995  Drexler  PorHou 11.7   5.9   8
1992  Ho Grant  Chi   14.1   6.7   0
1992   Nance    Cle   12.2   5.7   2
1991   Pippen   Chi   11.2   6.4   1
1986   Barkley  Phl   10.8   7.3   0
1978   Kareem   LAL   12.1   6.3   8

http://bkref.com/tiny/6DfJH
That final column is prior allstar selections. Some guys were not recognized as stars just yet.
* - Steve Francis leads the NBA in vorp.

In the last 4 seasons -- 104 selections + alternates -- median all-star WS is 9.3, vorp 3.8
Drummond should be starting.

[shadow]

Drummond doesn't look that great by vanilla RAPM. I haven't updated it since 1/22, but he was only 140th overall with splits of 87th on offense and 327th on defense.

His defensive rebounding numbers inflate his DBPM and consequently his VORP. It doesn't really pass the smell test to say someone has a +5.5 impact on defense when their team has a defensive rating of 107.2 when they're on the court and 99.9 when they're off the court.

[Mike G]

Yeah, I suppose the coaches who select the allstars can see that.
The Cavs' DRtg goes from 115 with LeBron to 105 when he sits. Doesn't affect their perception?

[shadow]

Well LeBron was the top vote-getter so it doesn't matter what the coaches think of his defense. We know the average fan voting on All-Star starters either doesn't care about defense or isn't great at evaluating defensive impact. I'm not denying his defense hasn't been awful though. He was 478th out of 488 in my latest vanilla DRAPM ratings.

[Crow]

LeBron is 360th on DRPM but the estimate is only -.80.

Less than 20% of players have estimates worse than -1 on DRPM. Only about 5% worse than -2. But on ORPM it is over 50% worse than -1 and almost 20% worse than -2. This doesn't surprise me but it does seem worth noting.

What are these rates on vanilla RAPM? How do they compare and if significantly different what to make of it?

Season is more than half over. Does Knarsu or any one have a grand table of well known and newer metrics side by side? That would be handy. Would aid comparisons, blends, questions, possible enhancements.

There are different dates for the RPM and RAPM estimates and they can be point in time volatility in each estimate for individuals. Has LeBron picked it up a bit on D? Were any of his estimates unusually high or low? A table displaying the point in time estimates would be interesting. Even though later estimates are larger sample and more likely to be accurate on average, I think seeing the variation, finding the average estimate or properly blended estimate and trend could help, especially to identify the most volatile, perhaps least reliable estimates. I have 2-3 point in time RPM estimates for this season saved. Might wrestle them together sometime later. Maybe at end of season. More points would be good though if anyone wanted to "wayback" recover them.

[shadow]

LeBron currently has a +2 DBPM. BPM is generally a good proxy for the box score prior in RPM. So that likely explains why his DRPM is a fair amount better than his vanilla DRAPM.

In DRAPM, 12.2% of players are worse than -1 and 2.2% are worse than -2.

In ORAPM, 13.2% of players are worse than -1 and 2% are worse than -2.

It makes sense that the rates would be lower in vanilla RAPM since everyone is regressed towards a 0 prior.

[Crow]

Thanks for reply.

RPM methodology is a powerful force. To get closer to "right" or not? User interest and ability to judge will vary. Not trying saves time and is easy but might not be the best "pro" response. Seems like it is probably pretty common though. Common among amateurs too.

Compared to vanilla RAPM, RPM distribution does stretch further away from zero especially on offense. Because of the mostly offense box score prior and / or the other adjustments.

[shadow]

Yeah, I don't believe vanilla RAPM is a better predictor than RPM. Testing by J.E. and others indicates that RPM is better in out of sample testing. I just like seeing what the vanilla RAPM ratings are since it doesn't seem anyone publishes them anymore.