RealGM

http://stats-for-the-nba.appspot.com/

It now just lists a bunch of teams and the RAPM for this season and the on-off.

What kind of information is missing that you are interested in?

It used to list the RAPM data for seasons from 2001-present. There is no way to get that information now.

blabla wrote:What kind of information is missing that you are interested in?

Everything, actually. All the data that was there that made the site useful historically is currently missing. I'm glad I copied as much as I did from the site while the other stuff was up, but I didn't get everything.

From what I can gather, J.E. has been tinkering with some new stats that may yield superior predicative power than RAPM.

http://www.apbr.org/metrics/viewtopic.php?f=2&t=8025&sid=2d8bfa203ce8d7cba2b3a54844de0fe4&start=30#p13810

I assume it'll all be back up soon

These numbers look changed drastically from last year? I.e. If anyone remembers the big Rose/Deng thread on the PC board, now we are seeing a complete reversal there, and many other players having gone up or down.

Edit - Reading the link above, obvious big effect from the box score influence. Perhaps not to be called 'the RAPM website' anymore.

Right, we're having this debate on a thread on the PC board, and I'm currently explaining my vehement disapproval on the APBRmetrics board. Just started having this discussion yesterday so I don't want to jump the gun, but I'm quite alarmed.

It's interesting because when I first got into all this stuff, I had major fights over on APBR where guys there just didn't see value in +/- stats. I had thought we'd gotten past these issues, but right now it seems to me like the guys who actually know enough to make the stats don't really understand why other people like using them so much. Kind of flabbergasted.

Doctor MJ wrote:Right, we're having this debate on a thread on the PC board, and I'm currently explaining my vehement disapproval on the APBRmetrics board. Just started having this discussion yesterday so I don't want to jump the gun, but I'm quite alarmed.

It's interesting because when I first got into all this stuff, I had major fights over on APBR where guys there just didn't see value in +/- stats. I had thought we'd gotten past these issues, but right now it seems to me like the guys who actually know enough to make the stats don't really understand why other people like using them so much. Kind of flabbergasted.

Any news about this? I need a fix of non-boxscore stuff.

So I guess RAPM is dead then?

No it's been brought back but it's not the same stat as before, it's been renamed as xRAPM and has become more of a statistical plus minus.

http://stats-for-the-nba.appspot.com

nikomCH wrote:No it's been brought back but it's not the same stat as before, it's been renamed as xRAPM and has become more of a statistical plus minus.

http://stats-for-the-nba.appspot.com

Yeah so it's dead.

nikomCH wrote:No it's been brought back but it's not the same stat as before, it's been renamed as xRAPM and has become more of a statistical plus minus.

http://stats-for-the-nba.appspot.com

How is that different from what it was previously? Is it possible to explain in layman's terms?

From what I understand he uses the box score to prior inform his RAPM (meaning something to base it off of) whereas before I believe it was non prior informed or RAPM informed. Either way the old version had absolutely nothing to do with the box score.

Doormatt wrote:From what I understand he uses the box score to prior inform his RAPM (meaning something to base it off of) whereas before I believe it was non prior informed or RAPM informed. Either way the old version had absolutely nothing to do with the box score.

That's half of what he's done.

He's taken the stat that you've described and blended it with the old "prior-informed" RAPM. That's what's listed on his site. It's essentially an average of a box-informed RAPM and the old RAPM

He says it's more accurate but the whole point of RAPM is to give us something outside of the box score so it doesn't make sense to me.

Doormatt wrote:He says it's more accurate but the whole point of RAPM is to give us something outside of the box score so it doesn't make sense to me.

I wouldn't say accurate, it's better in terms of making point margin predictions over the course of the season (which is generally how superiority is determined between these stats). But yeah, I have the same issues, I prefer the old RAPM.

For basketball analysis, it makes more sense to look at a non box-score figure in conjunction to box-score performance, and it sucks that we don't have any non-box-score +/- model publicly available right now. But I suppose that most of the stat guys don't really care about that, they just want to have the most accurate predictions.

Well maybe accurate isn't the right word but more reliable.

Also I think davidstern went back and did the calculations for the old RAPM, it's on google doc out there. I have it but I'm on my phone in class right now o_o

Doormatt wrote:Well maybe accurate isn't the right word but more reliable.

Also I think davidstern went back and did the calculations for the old RAPM, it's on google doc out there. I have it but I'm on my phone in class right now o_o

Yeah, rapcity posted it on the PC board a couple weeks ago I think, so I managed to save it

AussieBuck wrote:So I guess RAPM is dead then?

If by dead you mean "only available and being used by NBA teams", then yes, it's dead.

Slate actually had an article nailing the current phenomenon: NBA teams are hiring APBRmetrics folks - specifically +/- folks - faster than the internet seems to spawn them. Add in one crazy Engelmann literally not understanding the purpose of these stats as he produces them, and the public is totally cutoff for the moment.

Meanwhile of course, nba.com and b-r are light years ahead of where they were just a few years ago. I feel like the next reliably RAPM website we get might actually be one of those.

Doormatt wrote:Well maybe accurate isn't the right word but more reliable.

Also I think davidstern went back and did the calculations for the old RAPM, it's on google doc out there. I have it but I'm on my phone in class right now o_o

To get technical:

accurate & validity are synonyms: They ask about whether there's a bias in the data that moves the results away from the bullseye.

precision & reliability are synonyms: They ask about how consistent the data is.

The move from APM to RAPM was a move toward reliability and away from validity, which is why I never supported it as a one totally replacing the other, however the reliability jump was big, the validity fall seemed pretty small, and its usage was still completely orthogonal to the other tools available to us.

Engelmann's attempted move to XRAPM was again a move toward reliability and away from validity, but this time ruining the orthogonality relative to box score stats. I suppose what initially made me most annoyed is that Engelmann doesn't even seem to see his studies in terms of validity and reliability. He doesn't seem to even think about tradeoffs when designing a stat...and if your means of judging your stat is just well it collectively seems to make predictions that makes sense. No good analyst however is constrained to blindly use only 1 stat for all players.