| |
|
Review: Wages of Wins
That's a strange way to begin what is a book review of sorts, but I felt it needed to be said up front. When it comes to The Wages of Wins, the recent offering by sports economists David J. Berri, Martin B. Schmidt and Stacey L. Brook, a strange trend has been apparent to me and was confirmed by Malcolm Gladwell -- some of the harshest critics have yet to actually read the book.
It's certainly possible to understand the material in general terms without picking up a copy because Wages has drawn plenty of attention from the media, including a glowing review from the influential Gladwell in the "New Yorker," a review that was, ahem, less than glowing from 82games founder Roland Beech in the Journal of Quantitative Analysis in Sports and one somewhere in between by my 82games colleague David Lewin. Berri has written an editorial for the New York Times based on the material in Wages, and many of the general themes have been repeated on the authors' blog.
Naturally, as a basketball fan writing for a basketball Web site, my focus has been on Wages' discussion of basketball, which does make up the majority of the book. However, there is also work on baseball and one chapter on football, as well as a comparison of consistency across the three sports. Wages provides a good summary of work in the budding field of sports economics on topics such as competitive balance and how strikes affect attendance while breaking ground in these regards.
Wages promises to discuss serious topics in the casual, non-technical style popularized by the best-seller Freakonomics and succeeds in this vision; the book is a light, easy read, especially in comparison to other sports econ books I've read. At times, I actually found myself wanting more math. Details were promised on the Wages Web site, but I've yet to see them.
Wages wraps up boldly.
"Without statistical analysis, one cannot see how the actions the players take on the court translate to wins," the authors write in the final chapter. "One can play basketball. One can watch basketball. One can both play and watch basketball for a thousand years. If you do not systematically track what the players do, and then uncover the statistical relationship between these actions and wins, you will never know why teams win and why they lose. Staring at these players play is not a method that will ever yield the answers that the proper analysis of statistics will yield. And this is true if you stare for one day, or as we said, if you stare for a thousand years."
See? That's pretty bold. It's a direct attack on scouting and all non-statistical forms of analysis. To say something like that, you'd better make an argument that is so convincing as to be airtight. Unfortunately, on this count, Wages falls short. Too often, reasonable counter-arguments go unexplored as the authors seek to discredit conventional wisdom.
For example, Wages touts the strong relationship between scoring and salary in the NBA. What the authors fail to note, however, is that much of this relationship owes not to general mangers overvaluing scoring throughout the league so much as it does the relationship between salary and minutes played.
Naturally, points scored is a function of both minutes played and scoring rate per 40 minutes. So by using total points instead of scoring rate, the authors of Wages unintentionally introduce information about a player's passing, rebounding and defense.
Let us consider Minnesota shooting guards Trenton Hassell and Rashad McCants. Last season, Hassell outscored McCants, 710 to 627 (9.2 ppg to 7.9 if per-game stats are more your speed). However, no one would argue that Hassell was a superior scorer to McCants, who held an 18.4 to 11.3 advantage in points per 40 minutes. Hassell scored more points because his non-scoring abilities allowed him to play more minutes than McCants.
Amongst regular players in 2005-06, a player's per-40 minute scoring rate explained about 34.5% of how many minutes per game he received (correlation of .587). While scoring rate is a key factor in driving playing-time decisions, it is far from the only factor. As a result, it makes sense to separate scoring ability from playing time in a study seeking to explain salary (or All-Rookie team votes, another example used in Wages).
For my last column, I put together salary data that allowed me to approximate Wages' study linking player performance to free-agent pay. I looked at last summer's free-agent crop and 2004-05 statistics. The correlation between points scored and salary, .762, was strong, explaining 58.1% of the variability in the first-year salaries these free agents received. However, when we split this into minutes played and scoring rate, the correlation with minutes (.671) was much higher than points per 40 minutes (.532), which explains just 28.3% of the variability.
While I think there is likely a case to be made that scorers tend to be overvalued in the NBA, I'm not sure the way Wages presents the evidence makes a convincing argument.
The star of Wages is Wins Produced. In many respects, Wins Produced is similar to existing linear-weights measures, such as PER, but there are two critical distinctions.
One way in which Wins Produced differs from most existing measures is that it is significantly adjusted for team defense. John Hollinger has opted not to make this kind of adjustment with PER, arguing in his first Pro Basketball Prospectus, "That approach is incredibly crude; giving as much credit to Keith Van Horn as to Jason Kidd for the Nets' defensive strength [in 2001-02] just doesn't make any sense."
While there are reasonable arguments for and against a team defense adjustment, there is an obvious benefit for Wages. Making this adjustment allows the authors to show a strong relationship between the sum of players' ratings and a team's performance. This is a key argument Wages advances in support of Wins Produced.
The more crucial unique aspect of Wins Produced is that it essentially treats the ability to create shots - usually measured by basketball analysts as the percentage of their team's shots or possessions a player uses - as a non-factor. This is somewhat inevitable, because Berri's analysis started with the theory that the value of each statistic can be determined at the team level using regression analysis, then applied at the individual level. At the team level, creating shots is irrelevant.
Ironically, this aspect of Wins Produced theory is tripped up by a metaphor Wages introduces as an argument against scorers.
"Imagine that every team believed it needed to play its mascot to win basketball games," the authors write, drawing the comparison to teams believing their leading scorer is their best player. "If this were each team’s approach, half the teams would win with their mascots, the other half would lose. At the end of the season, one team with its mascot would be crowned champion, confirming the need to play your mascot."
Well, that is absolutely, unequivocally, inherently true of possession usage. Every league champion uses 100% of its own possessions; every last-place team uses 100% of its own possessions. If you are looking to determine the value of using possessions and creating shots, the team level should not be where you look.
The authors correctly argue that the NBA's Efficiency Rating, a linear-weights system in which every box score stat is valued the same, sets the bar too low in terms of scoring efficiency. A player shooting better than 33% on two-pointers or 25% from 3-point range will be shown as valuable to his team by shooting, even though both marks are well below league average (47.8% and 35.9%, respectively, last season).
Wages more accurately sets these levels at 50% and 33%, but this discussion raises an important question: Should the level of efficiency required to be an asset be the same for all players, regardless of how frequently they create their own shots? Wages answers yes, while the APBRmetrics community has generally answered no, as would conventional basketball wisdom.
As evidence of the theory that creating shots doesn't matter, Schmidt offered an entry in the Wages blog that noted, in terms of individual games, a positive correlation between shots attempted and field-goal percentage. However, this result is neither surprising nor inconsistent with a theory that players lose efficiency when they take on more possessions. When a player has a favorable matchup or simply a hot-shooting night, his teammates are likely to get him the ball, giving him more shots. He'll also likely play more minutes, meaning more shots if you don't adjust for shots per minute played.
Admittedly, evidence of the value of creating shots is scarce for the same reasons it is hard to prove this skill does not have value. One anecdotal way to explore the issue is to investigate how players play with and without a high-usage teammate on the floor. And, wouldn't you know, I happened to do this last season with L.A. Lakers star Kobe Bryant.
In the absence of wholly convincing evidence either way, I'm of the opinion -- though the authors of Wages, as well as some in the APBRmetrics community, would likely disagree with me -- that conventional wisdom should carry the day.
Let us not underestimate the importance of this distinction. Valuing the ability to create shots is the difference between Wins Produced ranking Allen Iverson the league's 227th most valuable player in 2003-04 and PER ranking Iverson 36th that same season. Iverson, always a lightning rod for controversy, has been the focus of much of the media's coverage of Wages.
By focusing on basketball, I do think the public perception of the book has been incomplete. Wages isn't really a book about how to evaluate basketball players; it is, at its core, a book that seeks to bring sports economics to the masses. It may not always have succeeded in shaking my conventional wisdom about the topics covered, but it certainly has caused me to revisit these beliefs and consider them in a different light.
But don't take my word for it.
For more discussion of The Wages of Wins, check out the APBRmetrics message board.
Mailing List
My writing for this site and SI.com follows no regular schedule. If you'd like to be notified when I've posted a new column, please e-mail me at kpelton@hoopsworld.com.
Kevin Pelton formerly wrote the "Page 23" column for Hoopsworld.com. He provides original content for both SUPERSONICS.COM and storm.wnba.com, where you can find more of his analysis of both the NBA and the WNBA. He can be reached at kpelton@hoopsworld.com.
Also see Kevin's previous columns for 82games.com:
|
| |
Copyright © 2006 by 82games.com, All Rights Reserved