From the folks at Talking Practice: [rquoter] A lot of people have been asking us on Basketball Twitter and APBR for a full list of RAPM values for the NBA, so we decided to oblige with the version that most people requested. The below numbers are created as follows: - This is RAPM-informed RAPM. - Seed year is 2006. All non-rookies start at {0,0} in the seed year. - Mean reversion of 15% is used between seasons. - An aging curve is applied to each player between seasons. It is currently specified as (27-AgeYrs)*.06 added as a linear term between seasons. - ROOKS take a prior of {-1.2, -1.2}. - XROOKS (rookies with substantial prior international experience) take a prior of {-0.6,-0.6}. - Priors for Rose, Bynum, and Frye were set at {0,0}. - All data parsed internally via NBA.com, crosschecked with 2 other sources, final verifications done via video review as needed. - We do not (and you should not) use these numbers for wagering purposes. We wouldn’t be giving them out to you if we did. - Changing any/all of the above parameters is reasonably trivial, so if the community desires any changes to the above, we can make some changes before the next run. We will likely update these about 2-3 times/week. [/rquoter] They say they plan to update it somewhat regularly, so this could be interesting to follow. Not sure if this is purely RAPM, or includes a box-score component like the xRAPM stat from stats-for-nba.com used to. Its not a particular precise way of measuring a player's value, but its one of the better methods out there for synthesizing all those lineup statistics that are so popular in the public domain these days. It incorporates RAPM from prior years as well, hence Asik (for example) still rates very well. Code: [B]Player RAPM RAPMO RAPMD[/B] James Harden 4.32 5.62 -1.31 Dwight Howard 2.41 -0.29 2.71 Omer Asik 2.34 -0.47 2.81 Patrick Beverley 1.48 1.09 0.38 Chandler Parsons 0.39 1.32 -0.93 Omri Casspi -0.29 -0.26 -0.03 Greg Smith -0.74 0.05 -0.79 Jeremy Lin -0.89 -0.14 -0.75 Terrence Jones -0.93 0.68 -1.61 Ronnie Brewer -1.63 -1.52 -0.12 Francisco Garcia -1.87 -1.28 -0.59 Donatas Motiejunas -2.02 -0.67 -1.34 Aaron Brooks -3.69 -0.88 -2.81
I want to re-iterate that these numbers aren't very good for players with very little playing time. Meaning that Jones and Beverley suffer a bit in terms accuracy. Everyone else on the team more or less fairly accurate. And depressing to see the negatives in the defense column. Morey really need to find another Battier.
The interesting number here to me is Dwight Howard's. He was a +5.5 last season in a "down year" and is significantly worse now. A 2.4 RAPM would make him more like a top 40 or 50 player than a top 10. Has Dwight Howard substantially regressed as a player since last season? Or is his effectiveness lower due to circumstances outside of his ability level?
If the +5.5 figure is from stat-for-the-nba, that stat (xRAPM) is calculated differently. It incorporates box-score based +/- model into it, whereas I think this one is purely RAPM.
Thanks for the info. Still, 2.4 doesn't look too good for a max guy and I would think it is a huge step down from his prior peak level.
FWIW, the same site has Individual-Player-Value, which ignores previous seasons with Harden 13th at 3.8 and Howard 16th at 3.6.
James Harden is number 2 (behind Lebron) in that offensive rating... at least he is really good at that!
Is this the same stats which ranks Lin in top 10 defensive rating midseason last year? Intuitively I'd move TJ up and move Asik down. Asik's stat is strange because his raw plus/minus should be pretty bad this season, so I'm not sure how they "adjust" it to be similar as DWH, given that DWH has slightly higher prior RAPM stat last year as well.
Interesting that Asik rates better offensively than centers like Bogut and Gasol despite their higher pedigree, though I'm sure that is almost entirely due to how he was made into a reasonably efficient player by James Harden. Also, this shows how much the Rockets missed Beverley (and Asik) over the past month explaining why they've struggled. It's encouraging that this team still sits with one of the best records in the league. Dwight Howard is paid too much money to have an impact that low (Roy Hibbert is rated as having a higher offensive impact and he's terrible). Hopefully, a lot of the regression of his RAPM is due to his down season last year and is merely "catching up."
I was looking forward to seeing some RAPM, but I'm no longer sure I can put so much weight in a system that rates Patrick Beverley and Jae Crowder by far and away as two of the three best second year players in the NBA along with Damien Lilliard. It also says that Harrison Barnes and John Henson are two of the three worst along with Tyshawn Taylor. Obviously, I'm probably missing a few names here so I'd more than welcome another set of eyes on this, but here is the list I came up with. 1. Damien Lilliard = 1.86 2. Patrick Beverley = 1.48 3. Jae Crowder = 1.30 4. Evan Fournier = 0.71 5. Jared Sullinger = 0.70 6. MKG = 0.42 7. Dion Waiters = 0.27 8. Anthony Davis = -0.03 9. Khris Middleton = -0.09 10. Draymond Green = -0.10 11. Darius Miller = -0.60 12. Austin Rivers = -0.65 13. Bradley Beal = -0.70 14. Andrew Nicholson = -0.71 15. Marquis Teague = -0.85 16. Terrence Jones = -0.93 17. Jeremy Lamb = -0.97 18. Quincy Miller = -1.03 19. Quincy Acy = -1.06 20. Kyle O'Quinn = -1.13 21. Perry Jones = -1.20 22. Jared Cunningham = -1.20 23. Andre Drummond = -1.21 24. Kendall Marshall = -1.22 25. Terrence Ross = -1.24 26. Pablo Prigioni = -1.33 27. Miles Plumlee = -1.63 28. Orlando Johnson = -1.70 29. Robert Sacre = -1.96 30. Mirza Teletovic = -1.98 31. Donatas Moiejunas = -2.02 32. Doron Lamb = -2.05 33. Robie Hummel = -2.14 34. Jeff Taylor = -2.30 35. Tyler Zeller = -2.36 36. Myers Leonard = -2.45 37. Thomas Robinson = -2.48 38. Tony Wroten = -2.66 39. Mike Scott = -2.74 40. Will Barton = -2.81 41. Maurice Harkless = -2.91 42. Harrison Barnes = -3.30 43. Tyshawn Taylor = -4.13 44. John Henson = -4.42
Harrison Barnes is playing 30 minutes a game and is putting up 10 points, 4 rebounds and 1.5 assists per game at below league average TS%. So it's not as if he is producing at a level that is contradictig that RAPM number. Barnes' DRTG is higher than his ORTG. Coincidentally, the Warriors fell off a cliff when he replaced Iguodala's minutes and subsequently took off when Iguodala returned. Perhaps Barnes is a more talented player than either Beverley or Crowder, but he doesn't fit in a team context as well as those two have up to this point. John Henson, on the other hand, plays a lot on the worst team in the league so it makes sense that he's probably not contributing much to winning right now.
I enjoy my advanced statistics just as much as the next guy, but this just doesn't seem to jive. I understand that there are many variables in this formula to come up with these numbers, but any formula that tells me that Marquis (D. League) Teague and Austin Rivers have been better players on the floor than Andre Drummond is a stat I need to ball up and throw away. That there are at least 43 more productive 2nd year players than John Henson, 22 more than Drummond, and 7 more than Anthony Davis. Seems like a case of over analysis gone terribly wrong.
This really comes down to what we mean by "have been better players". RAPM helps to quantify the extent to which players have contributed to their team's winning or losing, and is therefore dependent on how well the team takes advantage of his particular skills and hides his weaknesses. If by "better" and "worse" we're talking about basketball ability or individual statistical production, its possible that a "better" player isn't being used as effectively and thus has a lower RAPM. Or conversely a "worse" player is being used very well within his team's schemes and thus has a higher RAPM. The other thing to keep in mind is that RAPM is imprecise, and so if you're specifically looking for players who's ratings don't make sense you'll always be able to find some. As long as a large percentage of the time the ratings capture things pretty well, then it can still be useful.
Well, RAPM is not a measure of who is better or more productive. There are a lot of players more productive than Asik was last year, but we saw the sort of impact he had on the Rockets. Often production and impact coincide; sometimes they don't, which is the reason these sorts of stats exist in the first place. Anthony Davis, Andre Drummond and especially John Henson play on bad teams and are necessarily going to be punished by stats that attempt to measure impact. Of course, there also is the fact that RAPM uses a static prior for rookies which ends up accurate on average but will always produce outliers. Someone like Davis who was beyond his years even in college will suffer from that.
As far as I know, one issue with this type of stats is that there is always a "margin of error" (like the +/- 3.5% stuff you see with public opinion polls). These can be pretty large-- generally, I think the smaller the sample size, the larger the margin of error will be. Also, even though the formula used by the stats guys try to adjust for different factors, not everything can be totally accounted for. In any case, with any of these stats, I think it's one piece of info to think about and should not be thought of as the absolute measurement of a player's effectiveness. For front office guys like Morey, my guess is that he uses this information more to raise questions for further investigation (like "Anthony Davis seems really low. Is there something wrong with his situation that we don't know about or is the stat just wrong about him?") rather than using it to make final decisions (like "Anthony Davis' number is low. Ewww... do not want him at all! He sucks!"").
Totally agree. The starting point of this stat is team points scored and allowed while a player is on the court. Substitution patterns, random luck and not playing enough minutes per game or season make Landry Fields look like he's almost as good as Chandler Parsons even though he's only played 14% of this teams 1945 total minutes. Parsons has played 70% of the Rocket's 2079 minutes. Chandler Parsons 0.39 1.32 -0.93 DeAndre Jordan 0.38 0.99 -0.61 Marcin Gortat 0.34 -0.27 0.61 John Salmons 0.34 -0.44 0.78 Kenneth Faried 0.33 2.12 -1.79 Avery Bradley 0.32 -0.29 0.61 Landry Fields 0.29 0.51 -0.22 Chase Budinger is ranked near Beverley, but he's only played 7 games or 5% of his team's minutes this year. Clearly that sample size is too small and should be filtered out or at least be given an *. Plus/minus doesn't scale well by minutes played.
Here's an update for games through last night from the talking practice blog: Code: [B]Player RAPMO RAPMD RAPM[/B] James Harden +5.6 -1.4 +4.2 Omer Asik -0.4 +2.9 +2.5 Dwight Howard -0.2 +2.6 +2.4 Patrick Beverley +0.9 +0.6 +1.5 Chandler Parsons +1.2 -1.0 +0.1 Omri Casspi -0.5 +0.2 -0.3 Jeremy Lin -0.2 -0.6 -0.8 Terrence Jones +0.5 -1.6 -1.1 Donatas Motiejunas -0.7 -0.8 -1.5 Francisco Garcia -1.3 -0.6 -1.9 Aaron Brooks -0.6 -2.8 -3.3 Another set of "vanilla" RAPM numbers (from Jeremias Engelmann) has also been posted yesterday. It is based on past 3 years, with less weight given to past years (not sure exactly how he weighted it). http://stats-for-the-nba.appspot.com/vanilla_rapm/2012-2014.html Code: [B]Name OFF DEF OFF+DEF POSS[/B] Patrick Beverley +2.2 +2.3 +4.5 4072 James Harden +4.4 -1.8 +2.6 6073 Omer Asik -1.1 +3.5 +2.4 1214 Dwight Howard -0.8 +3.0 +2.2 6429 Chandler Parsons +1.0 +0.2 +1.2 6391 Jeremy Lin +0.0 +0.5 +0.5 4630 Francisco Garcia -1.6 +0.8 -0.7 3109 Terrence Jones +0.6 -1.5 -0.9 4665 Omri Casspi -2.1 +0.8 -1.3 3756 Donatas Motiejunas -0.4 -1.1 -1.5 1478 Aaron Brooks +0.6 -2.9 -2.3 2685 Here's a composite RAPM rating where I just took the average of the above two: Code: [B]Name OFF DEF OFF+DEF[/B] James Harden +5.0 -1.6 +3.4 Patrick Beverley +1.6 +1.5 +3.0 Omer Asik -0.8 +3.2 +2.4 Dwight Howard -0.5 +2.8 +2.3 Chandler Parsons +1.1 -0.4 +0.7 Jeremy Lin -0.1 -0.1 -0.1 Omri Casspi -1.3 +0.5 -0.8 Terrence Jones +0.5 -1.6 -1.0 Francisco Garcia -1.4 +0.1 -1.3 Donatas Motiejunas -0.5 -0.9 -1.5 Aaron Brooks +0.0 -2.8 -2.8
