Request a short overview of the current state of analytics

[Mike G]

What statistic is most likely to produce a ranking that would agree with the "common sense" opinion of the average basketball fan (as opposed to some advanced statistics that rank players like Draymond Green and Andre Iguodala very high).

Draymond and Iggy rank 2nd/3rd, on one of the best teams ever assembled, in their playoff run, in PER, WS/48, and BPM. That's some good corroboration.
They both also played a lot of minutes for a GSW team which was 2nd in ORtg and 1st in DRtg this year.
(In playoffs, they were 1st in both.)

Whatever you may consider "average fan" level of "common sense", there's no good reason to dumb it down to something that fails to consider team/opponent context. We could all see Messrs. Green and Iguodala guarding a wide range of opponents, holding their efficiencies down, while also working well on offense. It's hard to keep up with a team featuring multiple 2-way players like these.

[EvanZ]

What statistic is most likely to produce a ranking that would agree with the "common sense" opinion of the average basketball fan (as opposed to some advanced statistics that rank players like Draymond Green and Andre Iguodala very high).

Draymond and Iggy rank 2nd/3rd, on one of the best teams ever assembled, in their playoff run, in PER, WS/48, and BPM. That's some good corroboration.
They both also played a lot of minutes for a GSW team which was 2nd in ORtg and 1st in DRtg this year.
(In playoffs, they were 1st in both.)

Whatever you may consider "average fan" level of "common sense", there's no good reason to dumb it down to something that fails to consider team/opponent context. We could all see Messrs. Green and Iguodala guarding a wide range of opponents, holding their efficiencies down, while also working well on offense. It's hard to keep up with a team featuring multiple 2-way players like these.

Not too long ago you were pretty down on Draymond. He's come a long way.

http://www.apbr.org/metrics/viewtopic.p ... 374#p14374

[EvanZ]

How is the NBA.com PIE rating?


One last question. What statistic is most likely to produce a ranking that would agree with the "common sense" opinion of the average basketball fan (as opposed to some advanced statistics that rank players like Draymond Green and Andre Iguodala very high).

"Common Sense" = Points Per Game?

In all seriousness though, I think one could determine how fans view things by trying to regress All-Star voting on various features. My guess is one would find points per game, rebounds per game, and previous All-Star appearances as the top factors.

[DSMok1]

How is the NBA.com PIE rating?


One last question. What statistic is most likely to produce a ranking that would agree with the "common sense" opinion of the average basketball fan (as opposed to some advanced statistics that rank players like Draymond Green and Andre Iguodala very high).

"Common Sense" = Points Per Game?

In all seriousness though, I think one could determine how fans view things by trying to regress All-Star voting on various features. My guess is one would find points per game, rebounds per game, and previous All-Star appearances as the top factors.

I'd guess it looks like PPG+RPG+APG.

[Mike G]

Not too long ago you were pretty down on Draymond. He's come a long way.
http://www.apbr.org/metrics/viewtopic.p ... 374#p14374

Well yah.

Code: Select all

year    mpg   eFG%   TS%    PER  ORtg  DRtg  WS/48   BPM
2013   13.4   .354  .404    7.1    87  102   .028   -3.1
2014   21.9   .467  .498   12.7   102   98   .119    2.8
2015   31.5   .516  .540   16.4   110   97   .163    5.0

Who knows a good way to search for others with such immense improvements over 3 years?

[permaximum]

How is the NBA.com PIE rating?


One last question. What statistic is most likely to produce a ranking that would agree with the "common sense" opinion of the average basketball fan (as opposed to some advanced statistics that rank players like Draymond Green and Andre Iguodala very high).

"Common Sense" = Points Per Game?

In all seriousness though, I think one could determine how fans view things by trying to regress All-Star voting on various features. My guess is one would find points per game, rebounds per game, and previous All-Star appearances as the top factors.

I'd guess it looks like PPG+RPG+APG.

It would be "USG" only.

Still I remember average fan vote to predict wins for the next season did better than most analytics' predictions before... Like I said before, no all-in-one metric is probably better than MPG at player level (100% roster turnover). People who judge "players" by looking at WS, PER, BPM, RAPM, RPM are not really analytical. Until I saw a metric that's vastly better than MPG at out-of-sample prediction for a completely new team, my opinion won't change.

Oddly, we haven't seen these metrics' comparison to minutes per game at very high or 100% roster turnover. You can guess why...

[Crow]

I wouldn't use minutes per game by itself.

Looking at 41-60st highest / "best" on minutes per game (top 8-12% in league) here http://www.basketball-reference.com/pla ... y=mp_per_g 45% of the guys are rated below average on PER and ws/48, 35% on BPM. Is Jeff Green the 56th best player in NBA? I'll take PER, ws/48, BPM and RPM's answer over coach after coach.

[Crow]

Not too long ago you were pretty down on Draymond. He's come a long way.
http://www.apbr.org/metrics/viewtopic.p ... 374#p14374

Well yah.

Code: Select all

year    mpg   eFG%   TS%    PER  ORtg  DRtg  WS/48   BPM
2013   13.4   .354  .404    7.1    87  102   .028   -3.1
2014   21.9   .467  .498   12.7   102   98   .119    2.8
2015   31.5   .516  .540   16.4   110   97   .163    5.0

Who knows a good way to search for others with such immense improvements over 3 years?

that new app that waitlisted people for weeks / months last fall and created almost no lasting footprint to my knowledge? I might go back and try but my password stopped working or was lost.


anybody use it recently? https://www.statmuse.com/beta

Good idea but after a few tries I went back to manual production because of difficulties with syntax and getting exactly what I wanted the way I wanted it. I probably would be more gungho about drag n drop chart building than natural language based. I didn't give it much trial but to my knowledge nobody else either based on lack of running into examples after first few days. Another look / try is probably warranted but will it happen? I dunno. Not today.

[permaximum]

I wouldn't use minutes per game by itself.

Looking at 41-60st highest / "best" on minutes per game (top 8-12% in league) here http://www.basketball-reference.com/pla ... y=mp_per_g 45% of the guys are rated below average on PER and ws/48, 35% on BPM. Is Jeff Green the 56th best player in NBA? I'll take PER, ws/48, BPM and RPM's answer over coach after coach.

Well, there are two points (actually more but two major) to consider.

1. Blowouts' effect on MPG. Stephen Jackson and Draymond Green suffer from this.
2. The best player of PHI may get 42 mpg while Lebron gets 40 mpg in CLE. MPG is better used to compare players in the same team or teams with similar records and similar roster depth.

Despite these two huge cons of MPG, only RPM and BPM is better than it to predict next year and that's by a small margin. If I remember right when it comes to predicting actual point differential of all teams with previous season's metric values for each season between 1975-2014 MPG's RMSE was around 58 while xRAPM's 48 (for 2002-2014) and BPM's 53.

To make matters worse, at predicting net or wins for a completely new team MPG surpasses them all. But I must admit I tested it very roughly although I have a feeling I'm 100% right on this.

[Mike G]

A team made of 20-30 mpg guys from the Spurs and Warriors would not be the same as 20-30 mpg guys taken from the Sixers and Knicks.
Are you saying this could never happen in real life?

[permaximum]

I don't understand your question. I just pointed out the same statement that's it.

[Crow]

On average MPG on replacement is probably pretty good as you note and support. But individually it can be good to not so good. PER, ws/48 and BPM can be good to not so good at individual level, largely because of weak assignment of helping and hurting shot defense. RPM can be good to not so good. A weighted blend (including mpg in some fashion) might still be good to not so good, but it will probably be off by less compared to true talent or true performance impact on average, based on performance of blends in APBR prediction contests, sportsskeptic tests, xrapm over rapm.