14-15 RPM early estimates

[Crow]

http://espn.go.com/nba/statistics/rpm/_/sort/RPM

D Collison 16th overall. I Thomas a nice enough 14th pg. Estimate only half of Collison's (plus / minus affect of estimate errors.)

[Crow]

Comparing the known one metrics, most have K Thompson 20 something to 38th. RPM distinct on this one, ranking him 2nd.

Not as wide a spread with Duncan but still a spread of more than 20 spots from highest to lowest.

13 spots dif. for Boogie. 24 for LBJ. At least 12 for M Gasol.

There are biggest discrepancies I've found at this point. Just started looking. BOf vourse it is early and variances might shrink. But blends "better"? Still the question. Or, yeah, stop using some of the older ones (ws/48 and per) but that is not the whole issue.

[Crow]

Agreement on who is top 5-10% is probably the better story rather than making a lot of where they fall within it but it is a natural reaction. If RPM is mostly accurate on estimated impact to the level of say 5-7 tiers, that is pretty good. Much more than that may be optimistic.

[DSMok1]

Nice work, J.E.!

How much predictive accuracy did you have to sacrifice to make these reflect this season only, not including any prior season information? What impelled the change?

[Mike G]

Code: Select all

pos  %Min    ORPM   DRPM    RPM     WAR   war/48
C    .158   -1.17   1.27    .10     34.2   .071
PF   .212    -.22    .43    .21     47.8   .074
SF   .194    -.01   -.22   -.24     36.0   .061
SG   .222     .10   -.53   -.43     38.4   .057
PG   .214    1.08   -.56    .52     54.4   .084

tot 1.000     .021   .010   .031   210.8   .069

After about 25% of the season.

[Statman]

Comparing the known one metrics, most have K Thompson 20 something to 38th. RPM distinct on this one, ranking him 2nd.

Not as wide a spread with Duncan but still a spread of more than 20 spots from highest to lowest.

13 spots dif. for Boogie. 24 for LBJ. At least 12 for M Gasol.

There are biggest discrepancies I've found at this point. Just started looking. BOf vourse it is early and variances might shrink. But blends "better"? Still the question. Or, yeah, stop using some of the older ones (ws/48 and per) but that is not the whole issue.

Looks pretty darn close to my WAR, except I only use box score data, so my metric must be worthless (obviously can't compare to RAPM or metric "blends"). I can actually rate players from the beginning of the NBA, but all that matters is the here & now - we just gotta wait until we have enough data while pulling priors - that's the obvious way to go. Sure nice to know that IF my metric actually measures out great by Neil - it won't mean anything because obviously his testing method isn't correct (must be using in sample data to create metric, etc). Unless, of course, RPM or BPM come out on top - THEN obviously the testing is spot on.

Man, these metric debates are making me cynical.

[Crow]

It may get taken for granted because it is regularly done, but thanks for these stat comparisons Mike.

[Crow]

Statman, if RPM and WAR are close, at top or in general, that has meaning to me.

The old notion of having a website with all the one metrics side by side for players on a regularly updated basis doesn't seem unreasonable (assuming no copyright issues?) Hope to see it someday.

[colts18]

How does the xRAPM blend work? I don't think its been explained. Does it weigh +/- a certain percentage alongside the boxscore portion? Does it combine the two first then adjust by opponent/teammate xRAPM?

[Mike G]

Biggest losers by WAR/RPM. Thrown in are BPM and VORP for corroboration.

Code: Select all

. name           team   G    MPG   ORPM   DRPM    RPM    WAR    BPM  VORP
Luc Mbah a Moute PHI   19   27.0   -4.7    -.2   -4.8   -.80   -3.8   -.9
Zach LaVine      MIN   17   23.6   -1.4   -3.7   -5.1   -.62   -4.8  -1.2
Jabari Parker    MIL   22   30.0   -1.7   -2.1   -3.8   -.60    -.5    .9
Jason Smith       NY   22   15.8   -2.0   -2.9   -4.8   -.52   -5.1  -1.0
Nerlens Noel     PHI   17   30.8   -4.9    1.1   -3.8   -.47   -2.7   -.4

Dion Waiters     CLE   18   23.3   -2.9   -1.4   -4.2   -.45   -3.8   -.8
Nik Stauskas     SAC   21   13.9   -2.6   -2.6   -5.2   -.45   -2.6   -.2
Brandon Davies   PHI   20   19.0   -1.5   -2.5   -4.0   -.41   -2.8   -.3
Willie Green     ORL   19   15.5   -3.1   -1.3   -4.4   -.38   -6.9  -1.3
Austin Rivers     NO   19   22.1   -2.4   -1.4   -3.8   -.37   -3.4   -.6

. name           team   G    MPG   ORPM   DRPM    RPM    WAR    BPM  VORP
Quincy Acy        NY   19   22.2   -3.7    -.1   -3.8   -.36   -2.9   -.4
Doug McDermott   CHI   17   11.6   -3.1   -2.2   -5.3   -.36   -7.3  -1.1
Ke Caldwell-Pope DET   21   33.9   -1.3   -1.8   -3.1   -.33   -2.4   -.3
Gerald Henderson CHA   18   23.4   -2.7    -.8   -3.5   -.30   -2.6   -.2
Brian Roberts    CHA   20   18.6   -1.0   -2.6   -3.7   -.30   -3.1   -.4

D.J. Augustin    DET   21   22.6   -1.8   -1.6   -3.3   -.28   -4.0   -.9
Evan Turner      BOS   19   22.8   -1.3   -2.2   -3.5   -.28    -.6    .7
Anthony Bennett  MIN   19   17.7   -1.7   -2.0   -3.6   -.27   -3.3   -.4
C.J. Miles       IND   14   22.2   -2.9    -.8   -3.7   -.25   -5.1   -.9
Jarrett Jack     BKN   19   24.7   -2.6    -.6   -3.3   -.24   -1.9    .1

[DSMok1]

Statman, for what it's worth, I don't doubt that your WAR may be as good as BPM, I don't claim BPM to be perfection. I would like to see more testing. BPM/ASPM just happened to be picked up by Basketball Reference simply because it's been around a bit longer and has had more testing published, I guess.

Off topic here, though.

---

I wouldn't be surprised if RPM looks like a box score metric right now, because I expect that at this point in the season, it is basically a box score metric! We don't have a big enough sample size for on/off +/- data to have much weight in the blend.

--

The initial thread where J.E. introduced xRAPM is located at viewtopic.php?f=2&t=8025 I'm sure he's refined his methodology since, but it's a starting point.

[Statman]

I wouldn't be surprised if RPM looks like a box score metric right now, because I expect that at this point in the season, it is basically a box score metric! We don't have a big enough sample size for on/off +/- data to have much weight in the blend.

That would make a lot more sense - since we don't see any of the Nick Collison, Channing Frye, Amir Johnson, & Patrick Beverly types in the top 25 in this early release.

[Jinxed]

How are we to explain the surprisingly low ranking of the Bulls lauded frontline. Pau, Taj, and Mirotic all have mediocre rankings. Is that trio not playing up to snuff, despite what the eye test and traditional stats might tell us? Or is there some effect from having interchangeable players at the position so that there is not that much of a drop between line-ups and hence hard to parse out each individuals contribution. I notice that the three of them all of rankings bordering on 0.0..

[DSMok1]

J.E.... can you explain what is going on with Russell Westbrook's numbers thus far? I assumed that most of the current ratings are from box score data, but his rating indicates it may be otherwise.

[Crow]

On RPM: Parker, yuck, even more than standard for a rookie. Staukas? Yuck. McDermott, yuck. Noel, yuck. First one to neutral on RPM? In 3 years? More? BPM thinks best of Parker right now of 4. McD worst.

Someone on online said they were disappointed in Noel's offense so far. Yeah but what did they (Nd Hinkie) see in his college offense that they taught would be good in NBA? UK is only college team I watch and I didn't expect Noel to be much on offense.