Fair enough, but that's not all, it seems to think that Kobe Bryant sucks. I'm no Kobe fan, but the last time it had him even in the top 20 was the 2009-2010 season and never has him in the top 5, that seems a bit harsh.
Bball is all about size. Big men will always have the greatest impact. Reality sucks. Kobe is one of the most overrated players in this generation. In our society, experts do not make the most money. It's those who perceived' to be the experts make the most money. What people 'perceived' to be the best isn't necessary the best. It was Kobe's sponsors' job to make that perception amongst us. For a whole generation, they did exactly that by spending billions building up that perception.
Like I said, I'm no Kobe fan, but there were seasons where he was at least top 5 in the league, to suggest otherwise is a bit of a stretch. Also it doesn't have the leading scorer last year in the top 50 in the league. I know that Melo is the very definition of a chucker, but he's a top 50 player in the league.
A lot of time you can't really rely on stats to judge on a player's impact. Here are some examples: Ever since Carmelo came to New York in 2011, the Knicks have gone 11-10 in games in which he has not played. Top scorer or not, he had little impact. Lakers without Kobe had a record of 43-21, as of Feb 2012. That speaks volumely about what impact Kobe had. Before Dwight's injury, Magic was a solid 3rd in EC standing. After his injury, Magic went free falling to 6th and became the team that every other team wished to play in the playoff. He had a huge impact on the team.
To continue this conversation, which I find interesting: Nathan Walker at thebasketballdistribution.com made a projection for next season's RAPM based on player age, regression to the mean, and regression toward prior performance: http://www.thebasketballdistribution.com/2013/08/projected-xrapm-for-2014.html Somebody posted a 2013-14 projected Rockets win total by Kevin Ferrigan at NBACouchside.com based on each player's RAPM earlier in this thread. Ferrigan re-did his calculation based on Walker's projection and came up with this: http://nbacouchside.com/2013/08/26/re-doing-work-already-done-with-better-results/ A couple notable points: (a) The projection has Dwight Howard performing closer to his Lakers level than his significantly higher Orlando level (which would have given the Rockets 6 more wins or so). (b) The projection has Omer Asik playing 1748 minutes and Howard playing 2664 minutes on the season, this would require at least 476 minutes of overlap between the two guys (ignoring overtime). Not sure whether the Asik+Howard combo will work (which Ferrigan himself noted with some skepticism), but based on RAPM, the best way to increase Houston's win totals, assuming the spacing works acceptably on offense, is to find ways to play Asik more minutes because he's Houston's 3rd best player with an RAPM of +3.7 (21st best in the entire NBA!) and there is a big gap between him and the 4th best guy, Jeremy Lin at +0.6 (101st best in the NBA) and an even bigger gap between Asik and any of the PF currently on the team.
Interesting that Chandler isn't 3rd best. These measures don't seem to paint him in a very positive light.
Nice analysis Carl. 476 minutes a game works out to about 5.6 minutes a game, which agrees with my "sanity check" 5-6 minutes a game. My guess it that the coaching staff shoots for Howard at ~30 minutes a game and Asik at ~ 24 minutes a game. A great way to get them both into the playoffs healthy and rested. The chart you posted shows that the Rockets will be absolutely stacked at the SG and center positions. Makes perfect sense to me. I actually believe they will be much above average at the PG position if Lin can hit the low goal of playing like he did after the all star break. Also the increased minutes for Garcia will be interesting. If Howard can approach his 2011-2012 numbers there is a greater difference between Howard and the next best center (Gasol) than for any other position in the NBA. I do not view the gulf in between Harden and the next best SG (Wade, Kobe) to be near as wide. I view the projected wins of 58.72 as on the low side. It does not take into consideration the expected improvement of the young players Lin, Bev, Parsons, Harden, D-Mo, TJ etc. I also believe that Howard will be closer to his 2011-2012 numbers than to his Lakers numbers. But that is just a gut feeling based on my belief that he will be at full strength next season. I have several other points but this post is already long enough.
My thoughts on RAPM: 1. I think it is a pretty useful number and generally reflects the quality of a player's performance. If you look at where players ranks in terms of RAPM, for 85%+ of the players most people probably cannot make a well-supported argument that this guy is grossly better or worse than his RAPM ranking. The vast majority of the highest ranked guys like Lebron, Durant, Duncan, Chris Paul, etc. play a huge role on successful teams and the lowest ranked players are mostly guys who didn't really make any noticeable positive impact to his team's winning (Austin Rivers, Beasley, Ben Gordon). So, it's a pretty decent "rough approximation" of a player's value. 2. The fact there are some guys whose RAPM seem abnormally high or low to some or most people doesn't invalidate it as a useful tool. Like many statistical measurements, the RAPM is subject to sampling error (kind of like election opinion poll results come with a margin for error). It doesn't mean that you can't use RAPM, you just have to use it correctly. For example, the fact that Jeremy Lin had a +0.4 RAPM and Chandler Parsons had a -0.1 shouldn't be taken to mean that "Lin is better than Parsons," but should rather be understood as that these guys are most likely not that far from each other but each of their true abilities may well be +/- 1 or even 2 points away from their measured RAPM. We do know, however, that each of Lin and Parsons were most likely less impactful to the team than Omer Asik who had a +4.5 RAPM. RAPM is kind of a "rough" measurement. However, this doesn't mean that it isn't a useful measurement (like opinion polls used to understand the state of an election). 3. To the extent that there are some seemingly strange results coming out of RAPM, this doesn't mean we throw RAPM out of the window as a tool and this doesn't automatically mean that RAPM is right and conventional wisdom is wrong. It means that we should ask "How does this happen?" and look for answers using other sources of information (other forms of statistical data, game tape analysis, etc.). In fact, the existence of these situations is a big part of what makes RAPM results interesting and useful. For example, when Omer Asik was a fairly unknown backup C in Chicago, his 2011-12 RAPM was +2.8, 41st overall in the NBA. It would have been wrong to go "What, this scrub? Hahahaha!" As it turned out, Omer is in fact this good (and perhaps even better, his 4.5 RAPM in 2012-13 ranked 17th overall). On the other hand, I doubt that Morey relied on Asik's 2012-13 RAPM alone to decide on paying Asik $8.4M/year. Whatever measurements Morey used, the fact that they suggested a result that's contrary to popular opinion is what makes them useful as a tool for spotting opportunities in the free agent market. 4. Some of the interesting RAPM numbers that I noticed looking: A: DeAndre Jordan had a +3.1 RAPM in 2012-13 and +2.6 the year before. He gets a lot of crap for being overpaid and not that useful. The numbers could be wrong, but perhaps his shot alteration is actually more useful than most people think and he deserves more minutes. Maybe the Celtics were not stupid to ask for him in return for Garnett until the NBA blew up that trade. B: Ryan Anderson's RAPM was +4.4 in 2011-12 and +0.4 in 2012-13, with the drop mostly coming on the defensive end (+.1 to -2.9). RAPM is supposed to approximate a player's usefulness regardless of teammates as I understand it-- so this has to make us ask what happened. Were one or both of the numbers way off? Did Ryan Anderson become a worse player overnight? C. Jeremy Lin's RAPM in 2012-13 was +0.4, his RAPM in 2011-12 was +2.0 (probably higher during the peak of "Linsanity" before Melo returned). This isn't to say that Lin will for sure make a huge jump back to his peak level, but it does suggest that (a) Linsanity wasn't just empty stat stuffing, and (b) a significant improvement is at least possible. D. Taj Gibson's RAPM in 2011-12 was +3.6, his RAPM in 2012-13 was +2.5. The data suggest that Gibson may be an Asik-type sneaky-good player.
RAPM the overall number seem reasonable. But RAPM when divided into offense and defense can look utterly insane. I have never seen a metric that's so polarizing in the way it look at players. Case in point. Hakeem RAPM in 93-94: 6.0 4th in the NBA behind Deke, DRob, and Shaq. Okay. I don't like it as a Rocket homer. But I can accept that. Hakeem's offensive RAPM in 93-94: 0.3 For comparison with fellow players of >10,000 possessions used. Vladi Divac had 0.5, Horace Grant 1.2, Dennis Scott 3.7, Reggie Miller 4.2. Mark Price 5.8. To name one player from each position. Funny enough, Hakeem's finals counterpart Ewing had an offensive RAPM of -1.2(!). But overall RAPM of 4.6 that ranks him 10(again, reasonable rank).
In reality player fit matters, but the RAPM you cite associates a single performance measure with every player independent of whom they are playing with. Its possible that Ryan Anderson's skillset fit better (or was utilized better) on the 2011-12 team. Or, its also possible that the drop we're seeing is just due to random variation and Anderson was the same player playing at the same comfort level. That's one of the limitations of this form of RAPM (and pretty much every other player rating stat that's out there). Players can make each other better or worse depending on how well they "fit" together, but quantifying such effects is very difficult.
I wouldn't pay much attention to RAPM ratings from the 90s. They are approximations based on box scores.
Good point. How a player is used and the teammate around him changes his raw +/- and "adjustment" can't take care of this difference entirely. Fit matters and this is probably especially true for a guy like Ryan Anderson-- a role player with some pronounced strengths and weaknesses. I suppose what we can say is that Ryan Andersin can be an extremely effective player when playing next to guys who create space for his shooting and a defense that covers up his weak D. In other situations he may not make as positive an impact.
What about 2013? Lebron James defensive RAPM 1.6 Players higher than him in defensive RAPM: Kevin Durant, Mike Conley, Paul Millsap, Paul Pierce, Thadeus Young to name a few. And I purposely skipped true bigs here too.
Perhaps it shows that Lebron didnt excert his full energy on the defensive end during the regular season as much as he did in past seasons and he is so talented that he still ended up with a good defensive impact number, comparable to the output of other good perimeter defenders-- the small difference between Lebron's number and that of those other guys is well within the margin for error. The Heat were only ranked 9th in DRtg last season, worse than they have been in the past 2 years-- and they did not need to be any better. Being pretty much the guaranteed 1 seed in the East, it really doesn't benefit them to burn up too much effort on the regular season rather than saving their bodies for the playoffs a little.
On DeAndre Jordan: His issue is not his impact on the court, it's positive in general, roughly inline with what RAPM suggest, a useful body to stuff the lane who's not horrendous offensively. the problem is that he can't stay on it, because hacking strategies work way too well on him, the Clippers were one of the most under performing teams last year in terms of scoring differential versus actual results because of that (the other 2? the Thunder and the Rockets, I think we all realize that the Rockets late game execution was pretty bad last year.) On Lin: on the other hand RAPM does weight backwards, so it does somewhat suggest that Lin last year was mostly propped up a bit by the year before. which was obviously quite good, while Parsons was not quite as good due to general rookie issues. I think RAPM's weigh probably should take age / year in league into greater context. but that's not what it's entirely for. Even on pure unadjusted plus minus though, the two was virtually the same, however, so was Harden. while the best raw plus minus player was... James Anderson (among guys that played more it's Bev and Delfino. among starters the best was Asik by far.) RAPM is an interesting stat for sure, I feel it's better when we're talking about 5+ year vets where his skill level have stabilized, for younger player it's more random.
Look at Carl's table and pay attention to our new guys(Cassipi, Reggie...) making minimum salaries...Yes their RAPM numbers are surprisingly comparable to lots of well-known role players making millions in the league right now. Same story for Delfino, Garcia - before they even cracked any minutes with us last year, I read their RAPM/WARP numbers and felt good back then. RAPM maybe deceptively useful, and looks like Morey approves it with his moves.
RAPM is likely the most useful advanced stat that is publicly available.(actually temporarily unavailable last year) Also, you can bet that the Rockets incorporate some iteration of adjusted +/- into their player evaluations. Eli Witus, the Rockets current Vice President of Basketball Operations, was the former author of the blog Count The Basket, which explored adjusted +/- extensively. http://www.countthebasket.com/blog/ Like any metric, RAPM is just a point of reference. Like some have mentioned, a player's fit on a team may impact their rating, but also his role on the team. If a team used Shaq as a spot up 3pt shooter, or used Steve Kerr as your primary post up option, their RAPM will obviously be diminished. This is why Steve Nash went from a 1.5 RAPM in Dallas to a 4.8 RAPM in Phoenix. He was more effective in his expanded role. This is not to say every player would have a similar jump in effectiveness with a larger role, in fact that relationship is likely inverse. RAPM measures a player's impact on winning within the confines of the role he is given.
