Page 1 of 3

Well, well, well...

Posted: Mon Apr 07, 2014 5:23 pm
by schtevie
I was wondering when this would come to pass (and who would be involved):

http://espn.go.com/nba/story/_/id/10740 ... plus-minus

Let the controversies, real and imagined, ensue!

Re: Well, well, well...

Posted: Mon Apr 07, 2014 6:53 pm
by Mike G
RPM is a good name. Rolls off the tongue, kinda. Real plus-minus. Sounds very solid.
But it's really xRAPM?

We came up with the term "true shooting percentage" when we didn't have pbp to tell us whether it was truly true. And it turns out it wasn't exactly.

If RPM is Real, then Nick Collison is Really better than Tim Duncan; Patrick Beverly > Westbrook; Iguodala > Durant; Amir Johnson and Chris Anderson > all but a dozen players.

Carmelo's down there at #52; Paul George at 58 (after Jae Crowder); John Wall 60; Wade 70; Horford 80; Anthony Davis 88; Al Jefferson 99.

Re: Well, well, well...

Posted: Mon Apr 07, 2014 8:02 pm
by bbstats
My guesses

1) This is NOT multi-year and is therefore less stable, thus explaining low #s for certain players
2) ESPN / everyone decided to make it one-year so that it would be a little more concise and less confusing

Re: Well, well, well...

Posted: Mon Apr 07, 2014 8:26 pm
by J.E.
bbstats wrote:My guesses

1) This is NOT multi-year and is therefore less stable, thus explaining low #s for certain players
2) ESPN / everyone decided to make it one-year so that it would be a little more concise and less confusing
No and no. Still mulit-year

Re: Well, well, well...

Posted: Mon Apr 07, 2014 10:18 pm
by ilardi
JE and I may not fully agree on this, but my own feeling is that in most contexts the best means of comparing the relative overall impacts of two players is look at their related WAR (wins above replacement-level) ratings, also given on the main RPM page.

WAR can be thought of as reflecting the estimated impact of RPM x Possessions, by answering the counterfactual question: If all of this player's possessions this season had been taken instead by one or more players with the xRAPM impact of the typical "replacement-level" player (estimated by Daniel Myers at roughly -2.35 in his landmark analysis), how many wins would we expect this to cost his team?

Re: Well, well, well...

Posted: Mon Apr 07, 2014 11:35 pm
by EvanZ
I love it! Great job Jerry. :)

Re: Well, well, well...

Posted: Tue Apr 08, 2014 12:11 am
by colts18
J.E., if this is xRAPM why is there absolutely no mention at all of the boxscore component in the article? According to you, its 65% of the metric but its not mentioned in the article. If someone read the article, they would assume its some kind of NPI RAPM rating.

Re: Well, well, well...

Posted: Tue Apr 08, 2014 12:34 am
by talkingpractice
Obviously I think this whole thing is awesome.
colts18 wrote:J.E., if this is xRAPM why is there absolutely no mention at all of the boxscore component in the article? According to you, its 65% of the metric but its not mentioned in the article. If someone read the article, they would assume its some kind of NPI RAPM rating.
I have the same question (but I think it's more for SI than for JE). I was asked a lot today about (a) how much box was in it, and (b) whether it used a rookie prior. I wasn't totally sure what to say. I know that the ESPN audience in general doesn't care about either of these things, and so you may never clarify it there. So, can we just assume that the answer for this (and similar stuff) is identical to what we know of xRAPM?

Re: Well, well, well...

Posted: Tue Apr 08, 2014 12:46 am
by permaximum
This whole multi-year RAPM or prior-informed RAPM thing is wrong for the mainstream crowd. Just like ESPN, they'll take it and use it to determine how players behave in particular "one" season. Don't even get me started with the upcoming days of possibly endless "RPM" referrals for seasonal league awards.

The media and the target audience want non-prior-informed one-year RAPM. If you give them xRAPM or multi-year RAPM etc. they'll treat it as vanilla RAPM regardless.

Noisy or not if it's not vanilla RAPM, it will only lead to lots of incorrect assumptions.

Re: Well, well, well...

Posted: Tue Apr 08, 2014 2:55 am
by Kevin Pelton
permaximum wrote:The media and the target audience want non-prior-informed one-year RAPM. If you give them xRAPM or multi-year RAPM etc. they'll treat it as vanilla RAPM regardless.
And when they laugh it off due to the kind of fluky results we know are inevitable due to the noisiness of single-year RAPM, then what?

Re: Well, well, well...

Posted: Tue Apr 08, 2014 3:15 am
by HoopDon
And when they laugh it off due to the kind of fluky results we know are inevitable due to the noisiness of single-year RAPM, then what?
When they find out "RPM" uses previous season(s) performance, age, a "coaching factor", height, etc., they'll laugh it off anyway (I know as I've been pushing XRAPM on casual basketball forums for years).

NPI isn't super accurate, and its results can be a bit wonky (though the same can be and is being said of RPM), but its clean. People will know what it means, and how to apply it as a tool in the tool-box. People who have been using XRAPM for years still don't know how to use it, for some of the reasons listed above.

That said, this was a huge step for ESPN and NBA Analytics. More then anything, its nice to see J.E.'s work will get an even larger audience.

http://hoopdon.weebly.com/

Re: Well, well, well...

Posted: Tue Apr 08, 2014 3:42 am
by RoyceWebb
Jerry and Steve, separately and now together, have done fantastic work for years and I'm glad we can bring some of it to a bigger audience.

This board has been a great incubator for many ideas over the years, and for that I'm grateful as well.

This is Day 1 of the rollout. There will be more, with some of it geared to the broader public and some of it going into more depth about the specifics of these metrics. Both are important, though the former will often do more to inform people about advanced stats than the latter.

Your comments inform our thinking. Keep them coming.

Royce

Re: Well, well, well...

Posted: Tue Apr 08, 2014 4:12 am
by nbo2
WAR can be thought of as reflecting the estimated impact of RPM x Possessions, by answering the counterfactual question: If all of this player's possessions this season had been taken instead by one or more players with the xRAPM impact of the typical "replacement-level" player (estimated by Daniel Myers at roughly -2.35 in his landmark analysis), how many wins would we expect this to cost his team?
Estimating replacement level is extremely important and greatly affects the relative valuation of superstars vs. stars vs. starters etc. Daniel's analysis was a good start, but I think further analysis is needed. My guess is that the best player you can sign to a minimum 10-day contract is significantly worse than -2.35.

Re: Well, well, well...

Posted: Tue Apr 08, 2014 8:50 am
by Jacob Frankel
Massive step for NBA analytics, and can't wait to see what else you guys have coming over the next few days. Thanks a lot for investing in this, Royce et al.

Jerry, have the RAPMs from years prior to this one on stats-for-the-nba been updated with your new coach and score effect findings ? I'd imagine you'll be filling in previous seasons of data on ESPN fairly soon.

Re: Well, well, well...

Posted: Tue Apr 08, 2014 9:11 am
by J.E.
Thanks for the kind words, everybody.

For those wondering, it is essentially xRAPM without the coach adjustment. Rookies don't get any special treatment aside from the fact that the aging curve is part of the prior
Jacob Frankel wrote:Jerry, have the RAPMs from years prior to this one on stats-for-the-nba been updated with your new coach and score effect findings ? I'd imagine you'll be filling in previous seasons of data on ESPN fairly soon.
No. Everything but '14 on my site is 'old' (i.e. not reflecting the newest findings), so to speak