Advanced statistics are flawed, and people often use them out of context to support an argument which is what has led to the negative stigma advanced statistics now have on sports forums (except baseball forums maybe). Just because Morey uses advanced statistics doesn't mean everyone else can. Morey has access to billions of statistics that the public doesn't have access to, and he employs many great analytical minds who understand those statistics and how to use them, many of those great analytical minds likely create their own statistics. With all that being said, there are certain advanced statistics available to the public, and if used correctly some of them are somewhat useful. Some of you may already have heard of the statistic RAPM (Regulized Adjusted Plus Minus), it's like the regular +/- you can find in box scores except it tries to adjust for rotations. In other words, it tries to minimize the flaws of regular +/- which it does to a certain extent but not perfectly. It also recommends at least 60 games of data because small sample sizes exaggerate results (so for example don't bother looking at Kevin Love's RAPM, or bench players who don't get much), which is perfectly understandable although it still doesn't make the stat perfect. John Schuhmann was had an interesting talk with Erik Spoelstra where the usefulness of plus minus was brought up: I thought ES answered those questions perfectly. If you look at +/- and then go on to look at the different combinations, then +/- can get you to ask more questions and seek answers. HERE is the RAPM for 2012-2013 As you can see, Asik ranks not far below Harden suggesting what many of us already know--Asik was very, VERY, important to our team last season--almost as important as Harden. The biggest question I asked myself when I looked further down the list was "Why is Chandler Parsons so low?". Maybe it's because Harden often played SF while Parsons rests so Chandler wasn't missed. A back-court-wing of Beverley-Lin-Harden isn't a huge downgrade from Lin-Harden-Parsons. I'm just guessing though. So, if we use RAPM in context, is it useful? Or is it useless no matter what? Can I for example use it to support my opinion that Paul Millsap is the most underrated player in the league? Vote and discuss.
I think the Bev vs. Lin discussion re: who's a better defender has to be stopped...Lin's defensive RAPM and synergy stats are overwhelmingly superior to Beverly, who usually guards lesser talented guards.
It's popular to say that "advanced stats are flawed" because so many people are innumerate. If advanced stats are flawed then simple stats are flawed^2. The need to apologize for advanced stats is tiresome and a waste of time in pop sports forums where it falls on deaf ears. Many of Morey's views on "advanced stats" are published and provide very clear guidelines limiting how flawed you can claim they are if you agree that Morey knows what he's talking about, but most folks just continue to act as if he speaks nothing but air. The fact that, despite not having any history of coaching, or scouting, or playing basketball at a high level, Morey wins almost every trade he makes--and he makes a ton of them--should go a pretty long way toward mollifying the critics. But it doesn't. Anyway, RAPM is just a refinement of the least squares method for estimating the answers to a system of equations with too many variables, in particular the Tikhonov method of regularization: http://en.wikipedia.org/wiki/Least_squares#Tikhonov_regularization It's not arguable that it estimates players' historical affect on winning. The more data, the better the estimate. Of course, it's not as accurate at estimating players' potential effect on winning, which is bound to other factors like fit and scheme, e.g., Steve Nash was not as valuable to Dallas as he was to the Suns.
I should have said Morey wins every roster move he makes, since that captures more of the utility he gets from advanced stats.
As I recall, Lin's RAPM being higher is because the prior used for first year players is set very low (generally a sensible thing to do for making projections for young players, but can't assume this approach is accurate for all players). Beverly's defensive RAPM was actually better when you only consider possessions played last season, I believe.
On defense I never try to make too much assumptions myself. Besides using defensive rating, I basically go with what McHale's subsitution patterns look like. Who McHale likes to play. NBA coaches by nature are more prone to put in more defensive oriented players. And you can generally see who they trust to guard the opposing top players. Publically available advanced stats for offense IMO is much more accurate relative to actual player value and private stats used by stat-minded front offices.
useful for what? If you are Morey you need to understand why a person has a high RAPM not just that he has a high RAPM.
Beverley is the overwhelming superior offensive player based on RAPM and is slightly the better player based on RAPM. Can we cease pretending Lin is a good playmaker. / sarcasm
Cool thread. Never went in too much about Analytics. I think that something is greater than numbers. Hope Morey is not addicted to this thing...
Great post. RAPM is certainly bound to factors like fit and scheme, and Steve Nash in Dallas vs Phoenix is a perfect example. This leads me to thinking RAPM can be useful when voting for things like the MVP, because a player's RAPM is relatively reflective on how important a player is to his team, and a player's importance can be enhanced by the system he's in.
Advanced Stats are flawed. So are basic stats. The problems is, the creators of advanced stats promote their particular metrics as gospel, and many fans are pulled in by the ease of having an "exact" number to answer their question Case in point: All stats are useful, but only as long as you realize that they can never tell you the whole story.
I think a big reason why Parsons is ranked much lower than Asik is because Asik doesn't play with the 2nd unit. Parsons runs with the 2nd unit usually around mid 2nd quarter when harden was resting. Asik's RAPM is ranked high not only because he is a defensive anchor, but because he plays with very potent offensive scorers at all times.
The reason why Chandler is so low, or rather "average" on the list, is because he is an average player. The fact that he is talked about as a member of the "big 3" or a future all star speaks to how overrated he has become on this board. He is nowhere near that when it comes to his impact on winning. He is the same caliber player as Lin(slightly lesser actually), and people can't wait for the chance to throw $10M a year at him while reminding us everyday how overpaid Lin is. With that said, most rookies and second year players are net negative on the court. To already be an "average" player at this stage for Chandler is very promising. More importantly, he has shown an upward trajectory year over year, and even more drastic from the first half of last season, when he was close to a -2 RAPM. That however does not mean he is remotely close to a third "star", currently. The entire purpose of any "adjusted" +/- is to factor in the impact of the other 9 players on the court. This means who Parsons or Asik plays with, or against, has no bearing on their RAPM rating. For example, Mario Chalmers plays alongside 2 of the best players in the NBA, this is why his raw +/- is +13.2, the best on the Miami Heat. However, this number tells us absolutely nothing about the caliber player of Chalmers, nor his impact on winning. It is useless. RAPM on the other hand factors in the contributions of Lebron, Wade, along with the 2 other players on the court with Chalmers, as well as his 5 opponents his team is playing against. In the end his rating comes out to be +0.5, a very average role player, which makes a hell of a lot more sense.
Another great post! You are one of the most insightful members who were brave enough to predict a 45+ win last season (and actually bet $200 on it). I also remember Lin's RAPM was close to +1.5 about two months into the season. As it turned out, Parsons improved drastically to -0.1 while Lin regressed to +0.4. I am surprised that Delfino went from +2.0 to +0.0. The pace must've worn him down toward the end of the season. Matched him up on 4 didn't help either. Morey mentioned that Lin contributed the 3rd most wins behind Harden and Asik. There is a certain truth to it if he relied heavily on RAMP data. Bev played most of his minutes after the trade. If we take possessions into account, Lin did have the third most wins. I imagine it must be a distance third behind Harden & Asik and a lot closer to Parsons and Delfino.
Keep in mind last season was a season of transition and growth. You almost have to throw out the plus minus for this year as Lin, Harden, Parsons, Bev - where all adjusting to new roles and a new system.