Hollinger Rankings

John Hollinger ranks every player in the league based on a formula he has for projecting next season's likely PERs. It seems to me that taking the useful-but-in-some-ways-questionable tool of the PER as a ratings method and yoking it to the useful-but-in-some-ways-questionable tool of using similarity scores to project future performance is not the best way to get value for your number crunching buck. That Kobe Bryant (28.11), Dirk Nowitzki (28.20), and LeBron James (28.17) all did roughly the same on the PER metric last seasons seems like a valuable observation. The projection formula, by contrast, doesn't really add anything beyond what obvious qualitative nostrums ("LeBron's really young, he'll probably keep getting better"; "Kobe's been playing for a long time now, what happens if he loses half a step?") can offer us.

Delving into stats, Hollinger says that Agent Zero "might fare well if he played off the ball more, but the Wizards don't have anybody else to run the point." My strong recollection is that Arenas did play off the ball a non-trivial amount of time last year and that the Wizards do have someone else to run the point, namely Antonio Daniels.

Comments

This is totally beside your point, but the length of Kobe's career is basically a non-factor. If he'd gone to college, he'd have played nearly as much (fewer games, but much more strenuous practices, plus these guys love pick-up games). Kobe is 28 years old, with only minimally more wear than any other 28-year-old perennial All Star shooting guard. For comparison's sake, that would be 1992 MJ. Kobe is going to be very, very good for another half-dozen years or so.

By extension, I think breaking his collarbone was the best thing that ever happened to Adrian Peterson. This isn't the sort of injury that's going to be a continual problem (Charles Rogers is a remarkable outlier), as opposed to a joint or ligament injury, and it has probably saved him from 250 practice and game carries between now and season's end. Heck, I thought he should have sat out after the midpoint of last season, when he was nursing injuries and it was obvious Oklahoma wasn't a serious BCS contender.

Running backs' careers frequently end around age 30 because their bodies are beaten into submission, but nobody ever takes into account their collegiate beatings, only their collegiate propensity for injury. Football poses cumulative problems for the body, particularly that of running backs, which is why Ronnie Brown is uniquely suited for longterm NFL success.

Posted by jhupp | October 18, 2006 3:23 AM

Didn't Antonio Daniels play C3P0? Or was it R2D2? Talented guy.

Posted by Mudge | October 18, 2006 6:32 AM

Paul Pierce doesn't even crack the top 20? That can't be right.

Posted by P.B. Almeida | October 18, 2006 9:47 AM

but the length of Kobe's career is basically a non-factor.

I'm not sure I buy that. In college, you're not going to have back to back games. And you can't really treat practices and pick-up games the same way; less physical, and easier to slack off during either. Also, as I recall, the usage argument was first deployed to examine young pitchers, whose careers were supposed to last longer if they started pitching regularly in the MLB at or after 25. I'm not sure how baseball players are different than basketball players in your analysis.

We should start finding out soon. Garnett's turning thirty soon, and he heads the long line of straight-to-the-pros high schoolers.

Posted by SomeCallMeTim | October 18, 2006 9:56 AM

Hollinger projects Amare Stoudemire to be the fourth best player this year. Ahead of D-Wade.

Um... WHAT?

I think I like PER as a tool more than most people, but the projections are just crap.

Posted by Al | October 18, 2006 10:19 AM

i agree, for all the effort hollinger's put into devising this formula and analyzing why each player earns the score he does, the results aren't far off from a typical fantasy-league line. in other words, play a season of fantasy basketball, rank the players off the top of your head, and you'll end up with basically what hollinger has here. seems like a really round-about way of confirming what we already knew -- although we'll see if the system does a good job of pointint out underrated players, like jackie butler (now) of the spurs.

Posted by matt wright | October 18, 2006 10:34 AM

although we'll see if the system does a good job of pointint out underrated players, like jackie butler (now) of the spurs

Yeah - we'll see about Jackie Butler. But still, does anyone really think he'll be the 6th best center in the league this year?? Ahead of, say, Ben Wallace, Okur, Bogut, and Curry???

Another odd thing in the projections: Dwight Howard as only the 15th best PF??? Behind Channing Frye, Drew Gooden, Charlie Villaneuva, and somebody I've never heard of named Chuck Hayes??? Hollinger's on crack for that projection.

Posted by Al | October 18, 2006 10:49 AM

Agree with Al: there is no world in which anyone ever takes Drew Gooden (or Carlos Boozer or Channing Frye or...) over Dwight Howard. Does anyone think that even Isiah would pass up that trade? When your evaluation system produces insane results, it's time to rethink the system.

Posted by SomeCallMeTim | October 18, 2006 11:07 AM

Specifically on Howard, it's worth reading Hollinger's comment. It's not that PER gives Howard a low rating, it's that the similarity score system has a very pessimistic projection for Howard, suggesting he'll play substantially worse next year than he did before. Hollinger himself doesn't regard this as credible:

Howard is one of the most promising big men in basketball, so it might be a surprise to see his projected numbers for this year are so poor. Don't put too much stock in them --most players under 21 have very few comparable players to use for the projections, because leaving high school for the pros has been so uncommon. In this case, Howard's only comparables are Tyson Chandler and Eddy Curry, both of whom regressed in their third season.

I'd expect Howard's career to go in the opposite direction. His physical dominance isn't going to decline any time soon, and as his learning and skills increase he'll only wreak more havoc on opposing frontcourts. Provided the turnovers don't swallow him whole, Howard should be an All-Star this year and for many years to come.

So he doesn't actually think Howard is worse than those other guys.

Posted by Matthew Yglesias | October 18, 2006 12:05 PM

So he doesn't actually think Howard is worse than those other guys.

He just thinks that his projection system is bad. Which is kinda the point.

Posted by Al | October 18, 2006 12:11 PM

hollinger ranks duncan between shawn marion and chris bosh? that alone demonstrates how irrelevant these rankings are, even if they are somehow statistically sound.

Posted by MP | October 18, 2006 1:15 PM

He just thinks that his projection system is bad. Which is kinda the point.

Well, look, a projection formula that gives good results for 90 percent of players but offers wildly inaccurate projections for under-21 big men would be a very impressive achievement, notwithstanding some flaws. Someone just needs to remember to check back at the end of the season and see how well this system did.

Posted by Matthew Yglesias | October 18, 2006 1:30 PM

Baseball Prospectus does sort of the same thing and have a similar issue some players have lots of comparables and are easy to project, others are more unique, with Ichiro being an extreme example, I think they finally threw up their hands on him and simply said we have no way to predict what he will do. IIRC one thign they do is list a confidence % in the prediction based on how many comps a player has.

That all said, in general I find Hollingers predictions and system not all that usefull. Basketball is probably the most team-mate dependant of all sports vs Baseball which is the least. His rankings have too many WTF? results like Howard, Ray Allen and Paul Pierce just to pick 3 off the top of my head that seem ranked way too low.

Posted by Eric K | October 18, 2006 2:35 PM

if yao and tracey stay healthy, chuck hayes will probably be the rockets' third leading scorer from the assorted dishes and garbage clean up. dude has great hands and boundless energy.

Posted by paperpusher | October 18, 2006 2:46 PM

One thing about the PER ratings that is an artifact in the numbers is the very fact that it is per-minute based. There are some players that do very well in the 12-15 minutes a game that their coach finds for them and they become valuable players in such circumstances but can't anywhere near approach that kind of production given a different, usually larger, role. To extrapolate each player's production to a 48 minute average doesn't accurately rate that player's contribution since he very well might be only half as productive if give 25-30 minutes a game.

A per-minute based system cannot take this into account whereas a (competent) NBA coach probably does a very good job of it. One of the big factors that the coach can appreciate that the PER system wouldn't recognize (and thus can throw the results askew) are matchup differences between teams/players -- where an astute coach might be playing a certain player 25 min. against a favorable team but only 7 against another.

In this vein I'd say if your are even a reasonably astute observer of the game your own appreciation or ranking would most likely be at least as appropriate/accurate (and save you the 'insider' money you'd blow to get Hollinger's over rated system....).

Posted by Sigh | October 18, 2006 6:11 PM

Someone just needs to remember to check back at the end of the season and see how well this system did.

How well it did at what? Predicting a made-up number of questionable value? Or sorting the good from the very good? Because it doesn't seem to do anything like 90% of the latter.

Posted by SomeCallMeTim | October 18, 2006 6:13 PM