Monday, April 12, 2010

Book Review: Stumbling On Wins by David J. Berri and Martin B. Schmidt

Stumbling On Wins: Two Economists Expose the Pitfalls on the Road to Victory in Professional Sports by David J. Berri and Martin B. Schmidt

This is not a book for serious basketball fans. Because of that some of my criticism may be beside the point, possibly even unfair.

I am not an economist and am not qualified to evaluate the technical methods the authors use to measure wins.

I can evaluate the book's arguments which are, as a whole, unconvincing. That's not the same as saying the book isn't worth your time. I found the reading experience equally frustrating and thought-provoking. My resistance to the arguments helped focus my mind on what information* could be added to the boxscore to help compare the value of players.

*Current list: points created (including points scored on free throws) replacing assists, offensive and defensive possessions played placed next to minutes played, fouls drawn, separating defensive rebounds on free throws from total defensive rebounds, and separating technical free throw attempts from total free throw attempts to better measure each player's control over his own free throw rate.

The arguments are especially (and typically) unconvincing in the most provocatively titled chapter, "Defending Isiah." It's not actually provocative* as the defense of Isiah Thomas largely consists of claiming that the entire NBA overrates scoring rate and only James Dolan was willing to let his GM/Coach spend obscene amounts of money on otherwise unproductive players. The strongest defense is offered on page 29, showing that Isiah spent $10.5 million less than the authors' salary predictor model on the first year salaries of Stephon Marbury, Jamal Crawford, Eddy Curry, Quentin Richardson, and Zach Randolph. The conclusion:
[I]t looks like Isiah was making--according to the logic of the NBA's free agent market--very sound decision.
The first problem with that conclusion is the focus on just first year salaries. Curry's salary goes up $800,000 each year he's under contract** (and the Knicks will still owe him over $11 million next year). The total cost of Curry's contract was $56 million dollars. Furthermore, New York surrendered one first round draft pick outright (2nd overall, 2006) and gave Chicago the right to swap first-round picks (9th and 23rd, 2007). Acquiring Stephon Marbury also cost the Knicks two first-round draft picks. Berri and Schmidt do not account for the cost of those assets. Nor do they acknowledge that Curry's contract could not be insured. This seriously reduced his market value. Same with Quentin Richardson. The vast majority of the savings on the first year contract costs ($6.6 of the $10.5 million) were on the first year of two six-year, uninsured contracts totalling over $100 million combined. Isiah Thomas was not making decisions based on a solid understanding of the free agent (or any other) market.

*Leaving Isiah's destruction of the CBA unmentioned is provocative to this reader, but your reaction to that omission may be less visceral.

**Quentin Richardson's contract goes up $580,000 each year. Jamal Crawford's contract goes up $720,000 each year. That deal also ends after next season. Zach Randolph's contract goes up $1.3 million each year. Stephon Marbury's contract went up $1.8 million dollars a year.

The backbone of the book, the method for calculating the number of wins individual players produce demonstrates a similar disinterest in the complexity of basketball both in terms of the model's inputs and the certainty with which conclusions are drawn from its outputs. Most interesting to me, is the genesis of the model. Their primary impetus both for measuring wins and concluding that scoring rate is overrated* was a study of how NBA coaches vote for the All-Rookie Team. Where I might have concluded any demonstrable bias in that voting was down to coaches spending a tiny fraction of their time thinking about voting for the All-Rookie Team and thus giving undue weight to minutes played, draft position, and media attention, Berri and Schmidt created a model that supports (for better and worse) their theory that NBA professionals don't understand how basketball works.

*Which is not necessarily an incorrect conclusion if part of a more nuanced argument than offered in this book.

Are NBA decision-makers perfect? Certainly not. But I suspect that all NBA decision-makers consider far more information than do Berri and Schmidt in this book. There is no place for scouting* in the Wins Produced model. Many successful NBA teams do not dismiss +/- out of hand. From page 38:
[T]he purpose of tracking statistics is to separate a player from his teammates. A player's plus-minus, though, depends crucially on his teammates. If you happen to play with very good players, your plus-minus will tend to be high. If the quality of your teammates is low, your plus-minus will fall. As a consequence, the plus-minus statistic appears to contradict the very purpose of tracking numbers in sports. Rather than separate the player from his teammates, the plus-minus statistic defines a player by the quality of his teammates.
To me, is both blatantly incurious and rather missing the necessity of each basketball player interacting with four teammates at all times. Teams must explore the consequences of how players interacted or project to interact with each other. I'm sympathetic (in my ignorance) of the difficulties of modeling such interactions. That's not necessarily an argument for sticking with the boxscore stats, though.

*Even a Nick Fazekas partisan such as myself acknowledges the possibility that he is physically incapable of playing in the NBA though I wish he'd had the chance to contest that claim.

There is not, I suspect, an absolute answer as to the perfect balance between shot creation (or volume or scoring rate, if you prefer) and scoring efficiency. Each team needs players capable of both and very few players are capable of both. Though they speak admiringly of Larry Brown's entreaties to "play the right way" and that the 2003-04 Pistons won an NBA Championship without having a single player in the top 25 in the league in points per game that season, they're reticent on Brown's success in building a team in Philadelphia around Allen Iverson, and extremely high-volume, low-efficiency scorer. It would be interesting to know if Berri and Schmidt consider that (or Brown's acquisition of Stephen Jackson to improve Charlotte's offense this season despite Jackson using lots of possessions with a below-average TS%) an example of a great coach creating an unnecessary challenge for himself or an example of creativity and a deeper understanding of how basketball teams function.

Questions such as that aren't explored in depth in the book (equal space is devoted to studies of the NFL which are outside both my expertise and interest). Whether the authors have no interest in such questions or plan to address them at a later date, I do not know. If it's the latter, it's imperative they keep up with the work being done within the professional basketball industry. The value of creating and using an alternative to PER (without the scouting reports wherein John Hollinger provides context to the formulaic) reliant on traditional box score stats diminishes daily as work on adjusted +/-, defensive analysis, and video analysis progresses.

1 comment:

Zack Lerangis said...

While I also found it strange that many scorers who shot under 50% would be subtracting from their win shares, you have to understand this point: whenever a team has possession of the ball, that possession will either end in a FGA or a TO. If you acquire players who maintain possession of the ball (limit turnovers), then the overwhelming majority of possessions must end in a FGA. Essentially, someone has to eventually take a shot. The question is, then, if a team is full of players who shoot over 50%, but aren't necessarily scorers, will their FG% go down because there is not that "go to guy"? Basically, by being forced to take a few more FGA per game, will they have lower shooting percentages? Maybe, but I don't believe a dramatic decrease would be observed.