With this morning's report that Dwyane Wade is heading back to Miami with Chris Bosh in tow, the amazing free-agent scramble of 2010 is beginning to sort itself out. The Wade/Bosh report follows yesterday's news that the New York Knicks (the team I have rooted for since the mid 1970s), have signed Amar'e Stoudemire to a max contract worth $100 million over five years.
All of this is a prelude, of course, to the main event -- the LeBron sweepstakes. But what's been fascinating in watching this craziest of all NBA off-seasons unfold is that one player, David Lee, arguably one of the very best players available, has barely rated a mention. This omission is especially curious because in none of the commentary I have heard in the wake of the Stoudemire signing has it even occurred to anyone to ask whether trading Lee for Stoudemire even makes the Knicks a better team.
I know, at the start, that a large portion of the audience is going to write me off for daring to suggest that David Lee might actually be better than the likes of Stoudemire (and, while I am at it, perhaps Chris Bosh as well). But even if he isn't, there's almost no plausible way to look at their performances over the past three seasons and explain why Stoudemire just signed a $100 million contract, while Lee has gotten about as much attention as a car alarm in New York City.
So, before we get to Lee in particular, let's back up and talk briefly about NBA player evaluation and statistics. Thirty-odd years ago, Bill James launched a full-scale intellectual assault on the bastions of player-evaluation in baseball. The revolution he ushered in, known as sabermetrics has, after a long battle, over-turned a century of conventional wisdom about what statistics reveal the most about a baseball player's contributions to winning and losing.
James' approach, deepened and expanded by a generation of extremely smart, wonky men (and women) has now gone mainstream, both in the day-to-day language of baseball commentary, where discussion of things like on-base percentage and OPS have become commonplace and in the front offices, where a generation of general managers, led by Billy Beane, have made the newer approaches central to their assessments of talent. Not everyone buys the Moneyball approach, to be sure. But its basic insights are now part of the accepted parlance of the language of baseball conversation.
The NBA is, from this perspective, far behind major league baseball. There are a number of reasons for this, which I won't get into here, but there have been notable efforts in recent years to move behind the raw data of points scored, assists and rebounds to more systematic efforts to measure, statistically, a player's total contribution. John Hollinger at ESPN has developed perhaps the best-known of these new measures -- Player Efficiency Ratings -- (PERs). Dean Oliver, who now works in the front office for the Denver Nuggets is another statistical pioneer, having developed a system called wins shares. And David Berri, author of Wages of Wins and Stumbling on Wins has developed Wins Produced and related measures, also an effort to develop a statistically rigorous method for evaluating players. These methods, to varying degrees, have started to make their way into the larger conversation about player evaluation and the Denver Nuggets, who employ Dean Oliver and the Houston Rockets are two franchises with a reputation for using number-crunchers in figuring out how to fill out their rosters.
Bill James did a lot to debunk conventional wisdom in baseball discourse. But perhaps his most simple, straightforward contribution was to challenge batting average as the central statistic for evaluating players' offensive ability. Nowadays, though batting average still receives its obligatory attention, virtually any casual baseball fan (and even the players) are aware that things like on-base percentage, slugging, OPS matter at least as much (and likely more) as batting average does in understanding a player's contribution to his team's offense.
In basketball, as Dave Berri has pointed out repeatedly over the past few years in his books and blog, players continue to be evaluated, in general, based on their ability to score, despite the aforementioned recent breakthroughs in more comprehensive statistical evaluation. Berri has shown quite definitively that players who score the most points receive the most accolades (and the most money), notwithstanding the fact that there are plenty of NBA players with high scoring averages who don't actually help their teams win a lot of games. His metric, relative to others, gives significant weight to possession-related, as opposed to scoring-related, statistics.
Now, how do we know whether one statistical measure is a better guide to ranking players than another? One way to figure that out is by using a statistical method called regression analysis. Regression, in simplest form, is a way of figuring out to what degree one set of factors is related to another factor. For example, if you're trying to decide whether PER or Wins Shares (WS) or Wins Produced (WP) is the best way to determine how good a player is, you can use regression to see which system best explains how often teams win and lose. Since, in sports, the name of the game is winning and losing, that seems like a pretty good test of the value of a particular statistic.
And it turns out that Berri's WP does that better than any other system. Does it mean he's "right" about every player? Certainly not. All of these systems are based on probability -- the likelihood that one set of factors, or attributes or performance levels will yield a particular outcome, like winning a basketball game. There is no perfect way of doing this, and there certainly is no way to predict, with absolute certainty, whether a player's performance in one period of time (like the last three years) will tell us how he's likely to perform in another period of time (like the next three years). All of these systems are just best guesses.
But there is, as I've said, strong evidence that Berri's best guesses are better than Hollinger's or Oliver's.
This brings us back to David Lee.
Now, Amar'e Stoudemire does help his teams a lot. He's an excellent player, and all of the systems I mentioned above, including Berri's, regard him highly.
But how does he compare to Lee?
According to PER in 2009-2010, Stoudemire rates a bit higher than Lee. Stoudemire finished with a PER of 22.6, 9th best in the NBA (an average player's PER will be 15. An all-star level is about 20). Lee finished at 22.2, 12th best in the league. Over the past three seasons, Stoudemire has totaled 70.5, or an average of 23.5 per season. Lee's three year total is 59.2, just under 20 per season. So, per PER, Stoudemire is certainly the better player, though not overwhelmingly so (the average player will produce .100 wins per 48 minutes, or one-tenth of a win, which makes sense since there are ten players on the court at one time).
According to Oliver's win shares, Stoudemire's been worth a total of 31.4 wins to his team over the past three years, and per 48 minutes, about .197 wins.
Lee has been worth 26.7 wins to his team the past three seasons, at an average of .156 wins per 48 minutes.
So, according to Oliver, Stoudemire is certainly the better player, though, like Hollinger, his system sees very little difference between the two players over the past two years.
Berri disagrees, and quite emphatically. According to his metric, Stoudemire has contributed 28.4 wins to his team over the past three years. Per 48 minutes he's averaged about, .175 (as with wins shares, the average player would score a .100).
Lee, by contrast, has contributed 49 wins over the past three years to his (woeful) team, good for a per 48 minute rate of about .290. Incidentally, according to wins produced, Lee is substantially better than Chris Bosh as well. (According to PER, Bosh and Stoudemire have been more or less identical over the past three years, and according to win shares, Bosh and Lee have had little separation, with Stoudemire slightly ahead of both of them).
As I've mentioned, Berri's approach weights more heavily possession-related statistics - rebounding, blocks, steals and turnovers - than do Oliver's or (especially) Hollinger's. And Lee has been the superior rebounder, has a far better assist-to-turnover ratio than either Bosh or Stoudemire and racks up enough additional steals to make-up for his inferior shot-blocking (and neither Bosh nor Stoudemire is an especially good shot blocker).
In sum, Bosh and Stoudemire are superior scorers (not that Lee is a slouch in that department) and Lee does most other things better. It's also worth mentioning that Stoudemire, in particular, has had the benefit of Steve Nash at point guard these past few seasons, while Lee has had the (ahem) benefit of... Chris Duhon.
Of course, Heat fans are very happy to have Bosh teamed with Wade, as well they should be. And if LeBron announces tomorrow night that he's coming to NY, Knick fans, including this one, will be thrilled (and perhaps we'll find out that LeBron told the Knicks he wanted to play with Amar'e and not Lee. Who knows).
But on the merits, there isn't very good cause to justify the virtual writing out of the history books of Lee from the current discussion of top-flight free agents. He's been a monster on a crappy team, he's only 27, and has had less wear and tear on his body than either Bosh or Stoudemire (who've both been in the league quite a bit longer).
The statistical revolution remains in its infancy in the NBA, and players like David Lee remain under-valued (Carlos Boozer, a very similar player to Lee, has also been short-changed and overlooked, though not to the same degree as Lee).
# the data for PER and WS come from basketball-reference.com. For WP, the data are from Berri's wages of wins blog.
Jonathan Weiler's second book, Authoritarianism and Polarization in American Politics, co-authored with Marc Hetherington, was published in 2009 by Cambridge University Press. He blogs about politics and sports at www.jonathanweiler.com