The Superstar Thesis: Revised

[elrod enchilada]

Completley overhauled. Read in order.

elrod

2008. “The Superstar Theory: Revisited and Supercharged, Part 1,” NBADraft.net, 11 July. http://www.nbadraft.net/mcchesney005.html

2008. “The Superstar Theory: Revisited and Supercharged, Part 2,” NBADraft.net, 28 July. http://www.nbadraft.net/mcchesney006.html

2008. “The Superstar Theory: Revisited and Supercharged, Part 3,” NBADraft.net, 4 August. http://www.nbadraft.net/mcchesney007.html

[drza]

I like the premise of this work (i.e. a "superstar" is pretty much the price of admission to competing for a title). I like the conclusions of this work (the post-choosing "superstars" analysis is impressive). And I like the attempt to address some of the criticisms from the first trial, to make the argument less circular.

The main thing that I don't love is the method for choosing the levels of "superstar". It is mentioned in the intro of part 1 that APBR-metrics are getting much better, but in the end looking at the post-season awards was still deemed the best way to judge. The main problem with that, IMO, is that the year-end awards are heavily based upon how good a team is. The MVP vote is extremely skewed to the best team, as there is hardly ever an example of an MVP coming from a team that wasn't the first or second best in the league going back 25 years or so. That is what gives some circularity to the argument...MVPs (and to a slightly lesser extent the ALL NBA teams) are chosen from among the best teams in the league, and of course the best teams in the league also have the highest odds of winning a title, ergo the MVPs and ALL NBA-team members play on teams with the best chance of winning a title.

This weakness is addressed somewhat by looking over an entire career, as the Law of Large Numbers helps a bit with the circularity argument. Nevertheless, I think the whole ranking system would be strengthened quite a bit by choosing at least one APBR-metric measure and incorporating it. In a perfect world there would be a uniform APBR-metric measure that everyone agrees upon (PER comes closest, but the fact that it doesn't account for defense really weakens it IMO), but even barring that the measures that are out there are getting good enough to be used as a weighted portion of the superstar evaluation process. And since many of these measures really are more independent of team success and (in some cases) even strength of support, I think this objectivity would make the bottom line of this analysis much stronger.

[captain_cheapseats]

EDIT: oops, drza beat me to the punch. I basically agree with what he said: (1) the article was very interesting as always, and thanks for sharing/putting in all that hard work, but (2) you have a huge co-dependent variable problem that makes your results questionable.

[elrod enchilada]

Thsi is good criticism, that I have addressed the best I can with the resources and data available. Down the road we may be able to do better, but I fear it will be difficult to go very far into the past and generate the historical argument I make.

[drza]

Thsi is good criticism, that I have addressed the best I can with the resources and data available. Down the road we may be able to do better, but I fear it will be difficult to go very far into the past and generate the historical argument I make.

Well, a site like basketball-reference.com has PER data going back to the 50s. I don't love PER (like I said it needs more defense factored in), but at least it's relatively objective and looking at the All-time career leaders in PER passes the sniff test outside of a player like Russell (which is where the lack of D in the PER measure becomes more obvious).

http://www.basketball-reference.com/lea ... areer.html

The same site tracks defensive win-shares and defensive rating back to 1973. I notice you use a fudge factor for the All Defensive team (started in 1969( and the DPoy (started in 1982), so perhaps you could use a similar fudge factor for a defensive stat to correspond with PER in your objective measure.

My personal favorite APBR-metric stats are Wins Produced (wagesofwins.com) or the Roland Rating (82games.com), but neither one of them to my knowledge go back that far. I hear that adjusted +/- is also good, but I haven't looked into that one much at all.

[elrod enchilada]

Drza--

Two points:

1. My sense is that even if we located a very good statistical method to rank the players, we would pretty come up with the same list with only minor fluctuations. This is especially true at the top of the list, where most stars on championship teams are also found. All the statistical analyses I see pretty much come up with the same people as those who go all-NBA and get MVP votes.Hollinger's PER, for example, has all five first-team NBA players in his top 11, and three of the second team all-NBA are in the top 6. You have to get down to Bosh at 9th on Hollinger's list before you get someone who is not all-NBA. Do you think the gold medal category would substantially change, just being a fan of the game? I mean does anyone think as a result of some statistical system Alex English or Dan Issel is going to leap to gold medal status and Hakeem Olajuwon is going to get thrown overboard?

2. Recall, too, that my argument does not require that I be able to exactly rank order the best players in NBA history. I need merely be able to break out the best players by a non-subjective (i.e. not my opinion) measure applied the same to everyone and then clump them together into meaningful subsets. I make no clain this system has value beyond that. It is not perfect, and I address a lot of the problems in the piece. Since this is done over the course of careers this lessens the problem of one season or even two season oversights, if the goal is not a precise ranking but to create meaningful subsets.

So, bottom line, I see no evidence yet that the limitations in establishing a perfect list of best players undermines the general thrust or strength of my argument. If anything, it might just as likely strengthen it.

[drza]

1) I suspect that the final list would, in fact, be very similar to the one that you came up with. Which is, IMO, a good reason to do it...your conclusions would be stronger, and no one would ever be able to accuse you of circular reasoning again.

2) The reason to add the statistical measure is that I don't think you're using the term "objective" quite right. Objective doesn't just mean "not the writer's opinion", it means that it is independent of both opinion and other potentially mitigating variables. If the way that you define "superstar" is directly dependent upon how good a player's team is, you can't then make an objective statement that the team is championship caliber because of the superstar.

Your bottom line can be correct, but if you get there the wrong way it weakens your final statement. It was still impressive work, because this isn't exactly a peer-reviewed scientific journal that you have to meet the requirements for. But personally, I think that the premise and the analysis is worthy of being referenced as a legitimate source nationally. I just don't think it can be used as that type of source with that kind of circularity in the argument...even if in the end it still gives a correct answer.

So I hope you don't take it that I'm criticizing your work to be negative...just the opposite, I'd love to be able to site it in one of my blogs or my own articles. I'm just looking at ways that I think it could be better.

[ManchvegasBob]

Elrod -

All I can say is that I am leery of using any of the awards given to the NBA Players due to the bias that results from their award. I don't like compensating the PER with results from the DPOY and all-defensive team awards - I'd rather go straight to the statistics to temper PER, and that is found in 82games.com opponent production stats and the plus/minus stat of course.

Since this is a Celtics board and you also have some allegiance to the Celtics, you must realize that there may be a few flaws in rewarding the defensive efforts of offensive juggernauts like Ray Allen and Paul Pierce.

In http://mvn.com/nba-celtics/2008/04/13/pierce-talks-about-being-all-defensive/, I think the stage was set for Paul Pierce earning his first All-Defensive team assignment, but alas, that did not happen because the system award those that are defensive specialists that are biased because there offensive is so horrible. The exception is Kobe Bryant which got in with defense specialist because of his flair, but LeBron James, who is just as adept, has had no such luck because he's not in favor with the right.

How is the all-defensive team decided? From NBA.com:

The voting panel consisted of the NBA’s 30 head coaches, who were asked to select NBA All-Defensive First and Second Teams by position. Coaches were not permitted to vote for players from their own team. Two points were awarded for a First Team vote and one point was awarded for a Second Team vote.

I sense a potential for bias if this is indeed a popularity contest, which the blue collar guys that can't score well, compared to Paul Pierce and Ray Allen, seem to do well in. These guys include Bruce Bowen (1st - travesty), Raja Bell, Tayshaun Prince, and Shane Battier (2nd team - over Pierce - this also a traversty).

So in your first article http://www.nbadraft.net/mcchesney005.html, I can't imagine weighting who actually belongs on popularity contest of voting or based on sentimental feelings on what a player may have done in the past. I for one think that the Celtics with the 3rd best defensive team in the history of the league, only placing Garnett on the all-defensive team, is just hogwash, and I can't get beyond including any all-defensive team status as a metric in any ranking as a result.

Tell me I'm wrong Elrod . . .

- ManchvegasBob

[elrod enchilada]

You may be correct that the defensive votes are biased toward those who stink on offense. But that still does not prevent a lot of great offensive players from getting continual recognition for their D. Hell, Larrry made the team a couple of times. And I expect LeBron to get on the team within a couple of years. Chris Paul already made it.

The biggest problem with all-defense teams in my mind is that it waits a year or two too long before recognizing a player and then often keeps a great defender as year or two past their prime. By that logic it will tend to balance out.

As for an earlier comment, yes I know the definition of objective. It actually is a term I use and study a good deal in my "day job." I was using the term in a somewhat different sense, to suggest I was not simply providing my own opinion about players, which I do not think was inappropriate for the task at hand. I was not claiming, or at least I did not mean to claim, I was determining objective truth.