Regularized Adjust Plus-Minus (RAPM) numbers for Rockets 2013-14 are out

I was wondering if you were for real with the pretentious, condescending tone.

Thought you may be taking a stab at some JCDenton styled tongue in cheek schtick (which is pretty funny), but I guess not.

 

Fair enough. I do try to have all of my ducks in a row. I am really not good with opinions, but I can be pretty much correct when I am backed up by careful analysis. What comes off the top of my head is always suspect.

 

It had been explained it pretty well at one point, but I never saw exact detail. As you said, it brings in box score data. The box score data basically tries to see if a player's +/- is caused by the player contributions or a good situation. As I said, I use PIE to try to accomplish this in a qualitative way as RAPM has a hard time assigning causation for guys that play at similar times.

With the new player tracking stats that the NBA has made available this season, I suspect someone will come out with a stat even better than xRAPM. xRAPM is limited by the stats available at the time. I like the stats that show how effective a rim protector actually is and the stats that show whether a player is creating a lot of high percentage shots.

 

Hopefully. If it ever happens I would expect it to use SportsVU data and Synergy data. What we see as fans in SportsVU is a tiny slice of what NBA teams see. OK. I just stated the obvious.

 

Sorry, maybe I am misunderstanding your post. But isn't that link still pointing to regular RAPM? Not xRAPM?

 

As I've mentioned, I like the SportsVU data that is given on the stats.nba.com page. I think Synergy data is nice, but it can be subjective. I would not use Synergy data for a modified RAPM.

I tend to prefer stats to be separate though instead of the one stat fits all. Synergy is good at being descriptive of a player's game. One predictive element missing from xRAPM is fit. While it is great for stars, xRAPM may not be accurate for a player being placed in a role he is not good at. What will help the Bulls is not necessarily the same thing that would help the Rockets.

 

xRAPM has only been calculated at end of the year. The 2012-2013 season stats would be xRAPM. The 2013-2014 is a pure RAPM.

 

Am I missing something? At what point did JC Denton morph into JTR?

 

I believe you just dropped the mic.

xRAPM is gold, you say? How do you know?

No definition is available to check out its algorithms. But it's gold, right? I just have to believe in you, right?

Sounds cultish, but I've always been partial to cults. They're awesome.

That's a link to RAPM. Nothing new.

I'm guessing all of this stuff you post is off the top of your head, right?

 

When playing with the same 4 players, the Rockets offense is generally worse with Lin on the floor than with Harden, Beverley, Parsons, and Brooks. Rockets offense is better with Lin than with Garcia. I suspect the main reason for this is that Harden, Parsons, Beverley, and Brooks all can handle the ball and thus a guy that doesn't play well without the ball will struggle.

http://stats-for-the-nba.appspot.com/s4pm/2014/linje01_2014.html

 

This is all I could find that explains xRAPM:

http://hoopdon.weebly.com/1/post/2013/10/position-adjusted-xrapm-revised.html

So, the guy makes up percentages to follow a model. It still sounds like b.s. statistics. Don't hang your hat on it fellas.

 

xRAPM has been explained in the apbr metrics forum. I've posted links to the relevant thread in the past.

 

You have to take Hack-a-howard into account when talking about Howard's offensive efficiency.

 

You have not looked hard enough. xRAPM as an advanced metric ranks far above PER and WS, which are below WP, and somewhat better than ASPM. As far as a predictor of season NBA team rankings it was generally bested by Hollinger's rankings and Vegas odds, the two best in the business. But not by a lot.

 

There is one concept that people seem to miss in these stats. Sample size is very important. Some of the players don't have enough of a sample size to be compared to another player. Asik's defense isn't apples to apples to Howard because he isn't in the game enough to be compared fairly. In the case of Beverly, comparing his offense to Dragic is a bit odd not because of his playing time but because he doesn't get enough offensive looks since he is the last option on offense.

Thanks to Zergling for posting, good stuff.

 

No, xRAPM is RAPM with weighted percentages that float depending on what the creator of xRAPM wants to show. Hence, you don't know the formula. It's a big secret because there is no science behind it.

 

Absolutely true. While there has been speculation about xRAPM's underlying algorithms I am not convinced that they have nailed down how it does what it does. I call it a black box. But there is no arguing that it is not significantly better than RAPM and most other advanced metrics. And since those xRAPM figures have been published for the last 23 years of NBA data there is no argument that can be make about their validity. And many posters in advanced stats forums have taken a shot at how reliable they are.

 

They take shots at it because there is no science to it. "Black Box" is code for hidden. It's hidden because depending on what the creator is looking for, they change their percentages. It's bunk science. Glad you believe in it.

 

And I would also argue that none of the few attempts I have seen to recreate xRAPM contain "floating" adjustments.