Hollinger 2008-09 PER Projections

For those of you who understand John Hollinger's PER stat, congratulations: you're smarter than me.

Sort of. Hollinger, one of ESPN.com's better writers (and one of the original Basketball Prospectus wizards), uses a stat of his creation that, as a single number, is meant to represent a player's mean production per minute of time on the court. In other words, instead of looking at specific stats like points, assists, or steals, PER translates those numbers (via an immensely complicated equation that equalizes pace and minutes played*) into a single number that sums up a player's efficiency. In this particular system, 15 is the league average, with the league leader landing somewhere in the high 20s and low 30s.

So, with that in mind, check out the PER projections for 08-09 season.

If you are curious as to where he gets his dummy/projected stats to fill out the PER equations, he uses something he calls "similarity scores". Permit me to quote his explanation from a related article:

For each player, I use as a comparison the players from the past 20 years who are the most similar, based on age, height and stats over the past three seasons. Some players will have more comparables than others, depending on how unusual they are -- guys with freak heights (Yao Ming, Nate Robinson), freak ages (Dikembe Mutombo) or freak stats (Andrei Kirilenko) will have relatively few, while a more generic player like Al Harrington or Devin Brown could have over a hundred.

From that point, I see what their most similar players did a year later, and project those changes onto the stats of the player being studied. So, for example, the reason that J.R. Smith's PER is projected to rise sharply this year is because the most similar players also saw their PERs increase sharply at the same age; similarly, Andre Miller is expected to tank because a number of similar players hit the wall at his age.

Great, so maybe I shouldn't have drafted Andre Miller in my fantasy league... of course he does fill the utility slot as I also nabbed CP3 and Jose Calderon in an attempt to OWN the assists in my league. What uuuuuup...

But, seriously, despite CP's position in the numero uno slot, I have to say that Hollinger's stat is somewhat bothersome to me, and I've found that other writers agree (in theory... I'm not actually part of a circle that makes me privy to arguing with these people). For instance, Dean Oliver (author of Basketball on Paper) dislikes Hollinger's method since it produces a completely abstract number in place of quantifiable stats. Numbers that bear no relationship to accrued statistics (see: Quarterback Passer Rating) are difficult to digest, especially for the average sports fan. This is not to say that all sports fans are dumb, but rather that the complicated equations and seemingly arbitrary numbers that are the result hide the logic and brainwork put into their own creation. These stats force people to blindly accept what appears to be an arbitrary system of measurement in which one number is labeled "good", one "average", and one "bad". From there, you are left to extrapolate the rest.

Instead of player-based systems like Hollinger's (and there are many others like it), Oliver believes you're better off measuring team success than attempting to isolate one player's performance. He uses a simple method to calculate "efficiency" through an offensive and defensive rating.

This method produces a number that relates back to the most basic element of basketball: points. The resulting offensive rating reflects the amount of points a team will score for every 100 possessions (likewise, the defensive rating reflects how many points a team will allow from an opponent every 100 possessions). By bringing the possession number to a constant (100), we can effectively compare the run-and-gun offenses of D'Antoni and Nellie with the Mavs and Spurs more calculated (read: slow) approach.

Oliver then goes on to even deeper analysis using ratings to help mold larger extrapolations (at one point even calculating a win/loss record for individual players), but his basic argument is the same: basketball is a team sport with a myriad of player interactions on both sides of the court, and it is thereby almost impossible to truly calculate a single person's impact (though his formulas seek to do so).

This, of course, does not mean that Oliver's method is more correct than Hollinger's. In fact, Hollinger's method is a remarkable indicator of team success (i.e. despite Kevin Durant's gaudy point totals last season, he was never particularly "efficient", as he had to handle the load of his mostly incompetent teammates). It has also a reliable method for predicting rookie success in the NBA.

And so it goes. More and more "holy grail" of stats methods come out all the time. You just have to wade through it and make your own opinions. If nothing else though, it certainly does inspire a greater appreciation for the beautifully complex sport of basketball.

***

*Oh, and by the way, here's the PER formula in case you want to keep track at home:

uPER = (1/Min) * (3P + [(2/3) * AST] + [(2 - factor * (tmAST/tmFG)) * FG] + [FT * 0.5 * (1 + (1- (tmAST/tmFG)) + (2/3) * (tmAST/tmFG))] - [VOP * TO] - [VOP * DRBP * (FGA - FG)] - [VOP * 0.44 * (0.44 + (0.56 + DRBP)) * (FTA - FT)] + [VOP * (1 - DRBP) * (TRB - ORB)] + [VOP * DRBP * ORB] + [VOP * STL] + [VOP * DRBP * BLK] - [PF * ((lgFT/lgPF) - 0.44 * (lgFTA/lgPF) * VOP))])

where...

factor = (2/3) - [(0.5 * (lgAST/lgFG))/(2 * (lgFG/lgFT))]
VOP = [lgPTS/(lgFGA - lgORB + lgTO + 0.44 * 0.44 * lgFTA)]
DRBP = [(lgTRB - lgORB)/lgTRB]

then adjust for pace...

PER = [uPER * (lgPace/tmPace)] * (15/lguPER)

Done. Piece of cake.