(x)Regularized Adjused Plus-Minus Reference: http://stats-for-the-nba.appspot.com/ratings/2013.html http://www.swishappeal.com/2011/10/11/2484118/regularized-adjused-plus-minus-a-primer Original thread: http://bbs.hupu.com/5980295.html Sorry about no-translation, I only have access to image not excel datasheet. ================================== offense/100p defense/100p overall/200p total No. of Plays PG xRAPM sort by total No. of plays PG xRAPM sort by offense rating per 100 plays PG xRAPM sort by defense rating per 100 plays PG xRAPM sort by overall rating per 100 plays Compare to Jeremy's rating at Knicks Rating of current roster during last season
This is all I'm going to say about RAPM. Hakeem Olajuwon 1993-1994. Offensive RAPM 0.3 Vlade Divac 1993-1994. Offensive RAPM 0.5 James Harden 2012-2013. Offensive RAPM 6.0
Overall RAPM is much more useful than offensive/defensive RAPM because RAPM only accounts for how good the other team is overall, not just on offense or defense. What this means is that when centers like Olajuwon (or Howard) are in the game, the opposing team is not going to go small, they are going to make sure their defensive big men are in the game. For example, when the Lakers played OKC last year, Kendrick Perkins played big minutes because he's great at guarding Dwight. Perkins sucks at offense however. As a result, Dwight's defensive RAPM goes up and his offensive RAPM goes down, simply because when he is in the game a really crappy offensive player is in the game in Perkins. When Dwight comes out, the offense is much better because LA and OKC both went small, but the defense is much worse. Thus, centers that are great at offense like Howard and Olajuwon have artificially deflated offensive RAPM because good defensive centers are in the game while they play while teams go small when they don't play. However, they also have artificially inflated defensive RAPM. Thats why Olajuwon's offensive RAPM was only a .3 while Harden's offensive RAPM was 6. Overall however, Olajuwon had 5.7 defensive RAPM, giving him the 4th best RAPM overall in 1993-94 and a better RAPM than Harden.
eh, I'd rather use eye ball test. lin, why is the leading scorer at the pg position on pace to score 24 pts on you??? pg fail! we're trying to win hurrr.
I'm not as skeptical as some in here about RAPM, and I find it funny because didn't we just have a thread in the GARM gauging how many people believe in or use RAPM, and the VAST majority were for it? Yet when it is about Lin, the posters are all taking it w/ a grain of salt. Interesting...lol. Of course, like any stats, you have to look at it while looking at other stats (hoopdata, synergy, basketball-reference) to get a more accurate picture, and of course combine it with the eye test. I remember once having an irl discussion with a friend who believed Lin was terrible on defense (he isn't...he's like average), and how he can't stay in front of any PG. I argued that, while the latter may be true, that may be mitigated by the fact that very few PGs can stay in front of any PG in the league (i.e., comparing general speed vs. lateral quickness and the like). I like this because it sorta bolsters my point there, as far as defensive RAPM goes among PGs. His overall RAPM is kinda bad though. I mean, #20 out of 30 for PGs...essentially means he's below average overall as far as RAPM goes in the league as a starting PG, right? haha
Bingo. I think it is actually a really good stat, and it confirms a lot of what I see on the court overall. Side note: dang we need to find a way to get Lowry back.
Those numbers are completely wrong according to this link (website of the creator of RAPM): http://stats-for-the-nba.appspot.com/ratings/2013 Here's what it should look like: Player Offense/Defense/Overall Chris Paul 7.7/0.5/8.2 Russell Westbrook 5.3/0.1/5.4 Mike Conley 3.1/2.0/5.1 Stephen Curry 4.7/-0.5/4.2 Tony Parker 3.7/-0.4/3.3 Ricky Rubio 1.2/2.0/3.2 John Wall 3.2/-0.4/2.9 George Hill 2.3/0.0/2.3 Deron Williams 5.1/-2.8/2.3 Goran Dragic 3.1/-1.1/2.0 Eric Bledsoe 0.2/1.5/1.7 Devin Harris 1.5/0/0/1.6 Jason Kidd -0.3/1.1/1.4 Kyrie Irving 3.0/-2.0/0.9 Jrue Holiday 0.2/0.5/0.7 Steve Nash 2.5/-1.8/0.7 Damian Lillard 2.1/-1.5/0.6 Rajon Rondo 1.2/-0.6/0.6 Ty Lawson 2.9/-2.3/0.6 Mario Chalmers -0.3/0.8/0.5 Patrick Beverley 1.0/-0.5/0.4 Kemba Walker 1.0/-0.6/0.4 Jameer Nelson 1.7/-1.3/0.4 Jeremy Lin 0.6/-0.2/0.4 Isaiah Thomas 2.2/-2.2/0 Jeff Teague 0.7/-0.8/-0.1 Greivis Vasquez 2.2/-2.4/-0.2 Andre Miller 1.2/-1.4/-0.2 Jose Calderon 2.7/-3.1/-0.4
I understand perfectl what RAPM does. Which is why I find the whole breaking it down between offense and defense (a lot of LOFs have used this stat so say Lin is a great defender) to be utter BS. Yes, overall RAPM is actually quite in line with other metrics, both traditional and advanced, when judging a player. It also stacks up quite well in terms of the eye test. But breaking it down between offense and defense causes utterly ridiculous comparisons like the one I just stated. The issue with posting advanced metrics isn't posting them, but do so without explaining the reasoning and what they can do. And when you post stats like RAPM without explaining how to interpret them, how do you expect such stats to be treated?
Do you believe Vlade Divac was a better offensive player than Hakeem Olajuwon in 1993-94? Basically in your argument, the other person would say Hakeem was one of the best offensive players in the NBA, and you say he may not be better than Divac because Divac had 0.5 off RAPM while Hakeem only had 0.3.
Dwight Howard for example has the same offensive RAPM as Omer Asik, but that's not it saying that Howard's offensive game is the same as Asik's, however I'm also not sure how valuable it is for comparing different positions (it seems to be quite inadequate tbh), although in that sense it seems to give a picture on the most important skillsets for each position (e.g. elite defensive centers seem to always be more useful than offensive ones whereas elite defensive guards don't provide even half the benefit of an offensive one).
You know something is up if Lin has better defensive score than Beverley, while Beverley has better offensive score than Lin....
I hate (plus / minus) statistics. It is just a fundamentally flawed statistic. Its statistical value: 1 game = your kidding right?! 1 series = in one ear out the other 1 season = meh 10 seasons = ok I'll take a look a consider it with other relevant stats and environmental factors. 20 seasons = maybe there's a pattern emerging I can identify (as if I don't know this player well enough!) As a stand alone statistic it is weak to say the least. Combined with a comprehensive set of statistics it can be a factor (only with a large enough sample size of 5 to 10 years). I would like to see comprehensive univariate regression analysis performed to identify the explanatory value of "plus-minus" statistics to player performance for all players over the last ten years. I think it will come back with a poor Adjusted R^2 value. I suspect the correlation of multivariate regression analysis with other leading statistical indicators as variables is better.