Book Review: "The Wages of Wins" (Moneyball for the NBA)

[slamvangundy]

Wages of wins by Dave Berri (co-writers: Martin Schmidt and Stacey Brook) is an impressive attempt at applying the money ball ethos to basketball. Since reading the book, it has changed the way I watch and analyze basketball. I would say that the authors 75% agree with the conventional wisdom on which nba players are at the top or the bottom of the league. But they come up with a lot of counterintuitive points that are hard to dismiss---there writing is backed up with a lot of data and I think they are very measured in the way they make there arguments.

A few interesting points they argue:

1. Individual players tend to have the same output, regardless of who is on the court with them. Scrubs don't tend to have pretty constant efficiency/rebounding/assists and other statistical categories whether or not they are playing on the same court with superstars or other scrubs.

2. Nba analysts and fans largely base their evaluations of players based on points scored per game. Most of the top players in the game score a lot of points per game. But many of the top point scorers per game are mediocre or average in terms of the way they contribute to team wins.

3. Rebounding is vastly underrated in terms of a players contribution to team wins. Whether it is a point guard or a center, getting a few rebounds more than your positions average can make you a dramatically more valuable contributor to your teams winning percentage Getting a few less rebounds, can drive your value down from elite to above average, even if you are a point guard.

4. Players do not tend to improve past their third or fourth year. From year three till a player starts to wear out from old age, players contributions are pretty constant per minute. They test their ratings of players by projecting a player's career wins produced per minute. After a season if you multiply each player on a team's career wins produced per minute, by the number of minutes that each player played that season, it will statistically account for 95% of that team's winning percentage.

5. The top 20% of players in the league, tend to produce 80% of the leagues wins. The top three players on any given team tend to produce between 20 and 55 wins. Having a great trio with an above average supporting cast is essential to winning a championship.

http://www.amazon.com/Wages-Wins-Taking ... 0804752877

[Deeeez Knicks]

sounds pretty interesting. i agree with many of those points. might have to check that out.

[slamvangundy]

it's really a great read.

some interesting points about our very own knicks:

he ranks david lee one of the top 15 power forwards in the league:
http://www.wagesofwins.com/15PF0708.html

every season he has ranked eddy curry as one of the top 15 overpaid players in the league.
http://www.wagesofwins.com/Over41.html
http://www.wagesofwins.com/EddyCurry.html

jamal crawford last year was ranked one of the top 15 most overrated players in the league
http://www.wagesofwins.com/OverRatedPPG0708.html

how he ranked our team as a whole last year
http://www.wagesofwins.com/Knicks0708-76.html
Hint: our above average performers in '07 were lee, randolph and balkman. Most of our team are below average players and more than a handful have NEGATIVE value while they are on the court.

[noido]

A review of the book was published in the Journal of Sports Economics (April 2007) by Andrew Zimbalist.
The article is very critical of some of the techniques and assertions used in the Wages of Wins.
Unfortunately, Sage publishing will charge you $20 for short term access to the article. But if anyone is interested and has access through their college to the journal, or jstor, it is worth a look.

[sims]

i have some skepticism of win scores simply because they tend to really overrate hustle players with obvious limitations like paul millsap, carl landry, tyrus thomas, and our very own renaldo balkman. not that there's anything wrong with those particular guys or those types of players, just that it's misleading to think they can shoulder any more of their respective teams' load than they already do.

[Deeeez Knicks]

seems to overrate rebounds a lot. I mean rebounds are important but they have guys like Q, Zach and Lee rated pretty high because they are good rebounders.

Its tough to come up with accurate measure because things like man-to-man defense can't be totally measured in stats and that is half the game. There are a lot of intangibles in basketball.

[StarburyDynasty]

I really pushed WofW (not world of warcarft...) at the beginning of this past season and got ripped on hardcore. However, I may have deserved some of it, bc i've done research on it since then and found that WofW is really flawed--for example they arbitrarily value at least two of the statistics and their algorithm is kinda b.s.. I LOVE statistical analysis, but you have to take this with a grain of salt, and not just bc it says that AI is an average player even in his "prime," but rather there are plenty of flaws with the way wins are earned in their algorithm. there's plenty of in depth analysis/critiques of WofW online.

[j4remi]

Lamar Odom is listed 3rd to TD and KG. Q rich is our second best wins contributor...definitely some flaws there.

[DDV]

i can write a book on how to win in the nba in one page

get all time iconic big man surround him with one or two very good players=eventual title

or get three hall of fame to borderline hall of famers=eventual title

its not complicated

[slamvangundy]

A review of the book was published in the Journal of Sports Economics (April 2007) by Andrew Zimbalist.
The article is very critical of some of the techniques and assertions used in the Wages of Wins.
Unfortunately, Sage publishing will charge you $20 for short term access to the article. But if anyone is interested and has access through their college to the journal, or jstor, it is worth a look.

I'm not sure if I read that review, but for the sake of balance I'll point out one of the most convincing arguments against wages of wins that I have read. Almost everybody agrees that the margin a team beats other teams by, controlled for by possessions per game, is by far the best predictor of future winning percentage for a given team. When you sum up the player win scores on a team, you are inadvertently summing up to the margin a team beats other teams by, controlled for possessions per game.

That said, I think it is worth noting that many of the advanced statistical systems out there really seem to agree with each other in terms of who the overrated and underrated players are. I think it is noteworthy that WoW, PER and to some extent adjusted plus minus all tend to roughly agree with each other, and their assessments are often surprising about which players are underrated and overrated.

I think next season will be an interesting test case to evaluate some of Berri's hypotheses. You have the Ron Artest Move, Marcus Camby replacing Elton Brand in Golden State and you have Gilbert Arenas returning to the Wizards and the loss of Childress in Atlanta. Berri is going against the common wisdom by predicting that the Rockets fail to improve, that the clippers win 42 games, that the return of Arenas will not improve the wizards and that the loss of Childress will hurt Atlanta.

Last year Berri was derided by many for predicting a boston title, for predicting a 20 win season for the Knicks, for predicting Philly's playoff appearance, etc. While some of the criticisms of him are worth evaluating, and while no model will tell you the whole story, I think Berri is definitely onto something. At the very least, his book is very worth a read.

[noido]

I'll list the main arguments

- The assertion that the NBA is uneven because basketball requires tall people ignores the fact that until 2002 at least the NBA had a porous salary cap and no revenue sharing

- The assertion that labour action does not cause an attendance drop off is incorrect, WoW comes up with this by comparing 1993 and 1996 baseball attendance, when they should be comparing the 1994 and 1995 seasons (where a 20% drop off occured despite robust economic conditions)

- The assertion that each all star vote for a franchise's player equals 16 cents in revenue is not properly explained. It is not made clear if an OLS or Tobit model is used. The website does not have detailed explanation of the analysis despite the assurance in the book (this would hinder critical examination of the methodology used)

- The underlying data is problematic. It is not clear when payroll is calculated (beginning, middle or end of season) which causes difficulty in finding a relationship between payroll and performance. Revenue data appears to be from Forbes, but these estimates can be substantially off the mark

- Claims that scoring is overrated by GM's which is why there is low correlation between payroll and performance, yet ignores rigid labour market rules and less predictable team chemistry. In any event if GM's did this, one would expect a labour market correction, which has not occurred.

- Asserts that stars dont perform better in the playoffs by examining overall productivity during a game, but does not look at intra-game performance during crucial periods.

That is all for the NBA criticisms

-

[god shammgod]

2. Nba analysts and fans largely base their evaluations of players based on points scored per game. Most of the top players in the game score a lot of points per game. But many of the top point scorers per game are mediocre or average in terms of the way they contribute to team wins.

sounds like a good read. people do tend to look at points per game as an evaluation of a players's performance. people will say jamal crawford is a 20 point scorer but will ignore his horrible percentage. it doesn't surprise me this guy thinks he's one of the most over rated players.

[mjhp911]

it's really a great read.

some interesting points about our very own knicks:

he ranks david lee one of the top 15 power forwards in the league:
http://www.wagesofwins.com/15PF0708.html

every season he has ranked eddy curry as one of the top 15 overpaid players in the league.
http://www.wagesofwins.com/Over41.html
http://www.wagesofwins.com/EddyCurry.html

jamal crawford last year was ranked one of the top 15 most overrated players in the league
http://www.wagesofwins.com/OverRatedPPG0708.html

how he ranked our team as a whole last year
http://www.wagesofwins.com/Knicks0708-76.html
Hint: our above average performers in '07 were lee, randolph and balkman. Most of our team are below average players and more than a handful have NEGATIVE value while they are on the court.

I would agree with most of those findings.

Almost everybody agrees that the margin a team beats other teams by, controlled for by possessions per game, is by far the best predictor of future winning percentage for a given team. When you sum up the player win scores on a team, you are inadvertently summing up to the margin a team beats other teams by, controlled for possessions per game.

As fun as run and gun is to watch, I still firmly believe that you win games in the postseason by your defense, and controlling the number of possessions each team gets (to the extent possible). Which is why I've always thought Phoenix was just eye candy, a regular-season wonder. But I'm willing to see where 'Antoni can take us, and I'll support him early. I wanna' see progress though...

[rpa]

3. Rebounding is vastly underrated in terms of a players contribution to team wins. Whether it is a point guard or a center, getting a few rebounds more than your positions average can make you a dramatically more valuable contributor to your teams winning percentage Getting a few less rebounds, can drive your value down from elite to above average, even if you are a point guard.

I stopped after reading this part. Individual rebounding (in my mind) is actually an incredibly OVERRATED statistic. Team rebounding is important but it's just like team defense: just because you're a great defender doesn't mean your team's defense will be great (and vice versa). Rebounding is so dependent on many things; just one example:

consider a team (like the Suns) where the guards (and small forward) are told to break up court after closing out on a jumpshooter. This would not only limit their chances for rebounds (and thus their total) but it would also artificially increase the likelihood of one their frontcourt players grabbing a rebound.

[slamvangundy]

3. Rebounding is vastly underrated in terms of a players contribution to team wins. Whether it is a point guard or a center, getting a few rebounds more than your positions average can make you a dramatically more valuable contributor to your teams winning percentage Getting a few less rebounds, can drive your value down from elite to above average, even if you are a point guard.

I stopped after reading this part. Individual rebounding (in my mind) is actually an incredibly OVERRATED statistic. Team rebounding is important but it's just like team defense: just because you're a great defender doesn't mean your team's defense will be great (and vice versa). Rebounding is so dependent on many things; just one example:

consider a team (like the Suns) where the guards (and small forward) are told to break up court after closing out on a jumpshooter. This would not only limit their chances for rebounds (and thus their total) but it would also artificially increase the likelihood of one their frontcourt players grabbing a rebound.

Except that Suns guards and forwards are among the league leaders in wins per minute. It is likely that the increased field goal percentages those players realize by breaking across court early outweigh their supposed lack of rebounding.

Marion and Steve Nash are both rated among the elite of the elite of NBA players.
Grant hill rated as an above average player.

In the the wages of wins system (which is over 95 percent correlated with team winning percentage), players can compensate for being out rebounded, by scoring a lot of points at high efficiency and having low turn overs. Don't be so knee jerk. Read the book, and understand berri's argument before you reject it. You might learn a thing or two!

[slamvangundy]

3. Rebounding is vastly underrated in terms of a players contribution to team wins. Whether it is a point guard or a center, getting a few rebounds more than your positions average can make you a dramatically more valuable contributor to your teams winning percentage Getting a few less rebounds, can drive your value down from elite to above average, even if you are a point guard.

I stopped after reading this part. Individual rebounding (in my mind) is actually an incredibly OVERRATED statistic. Team rebounding is important but it's just like team defense: just because you're a great defender doesn't mean your team's defense will be great (and vice versa). Rebounding is so dependent on many things; just one example:

consider a team (like the Suns) where the guards (and small forward) are told to break up court after closing out on a jumpshooter. This would not only limit their chances for rebounds (and thus their total) but it would also artificially increase the likelihood of one their frontcourt players grabbing a rebound.

If that were the case, individual players rebounds per possesion would vary wildly from year to year, depending on what kind of system they were playing in. However, most player's rebounds per possesion tend to be fairly constant. Dean Oliver who wrote basketball on paper addresses this issue.

[slamvangundy]

I'll list the main arguments

- The assertion that the NBA is uneven because basketball requires tall people ignores the fact that until 2002 at least the NBA had a porous salary cap and no revenue sharing

- The assertion that labour action does not cause an attendance drop off is incorrect, WoW comes up with this by comparing 1993 and 1996 baseball attendance, when they should be comparing the 1994 and 1995 seasons (where a 20% drop off occured despite robust economic conditions)

- The assertion that each all star vote for a franchise's player equals 16 cents in revenue is not properly explained. It is not made clear if an OLS or Tobit model is used. The website does not have detailed explanation of the analysis despite the assurance in the book (this would hinder critical examination of the methodology used)

- The underlying data is problematic. It is not clear when payroll is calculated (beginning, middle or end of season) which causes difficulty in finding a relationship between payroll and performance. Revenue data appears to be from Forbes, but these estimates can be substantially off the mark

- Claims that scoring is overrated by GM's which is why there is low correlation between payroll and performance, yet ignores rigid labour market rules and less predictable team chemistry. In any event if GM's did this, one would expect a labour market correction, which has not occurred.

- Asserts that stars dont perform better in the playoffs by examining overall productivity during a game, but does not look at intra-game performance during crucial periods.

That is all for the NBA criticisms

-

Great post noido!

I am definately of the opinion, that Berri is far stronger on analyzing player productivity than the economics of the nba. Your criticisms about salary determination and attendance are definitely valid. However I think there is great value in a system that accurately rates teams as a sum of their parts.

Even if Berri uses somewhat tautological reasoning at teams, that the value of the players on a team corresponds with team wins is a big breakthrough for player rating systems. The knicks have many players who have been double digit scorers/rebounders/assist men in their careers, but as a whole we are garbage. To me, a system that can accurately predict a teams winning percentage, based on the players on that team, provides a lot of valuable insight into the nature of winning in the nba.

Is there room for more nuanced analysis? Yes. But what berri has laid out is the best system currently available to the public and I will take that over cliches about team's "waiting for the right chemistry" and players' "clutchness" any day of the week.

[slamvangundy]

it's really a great read.

some interesting points about our very own knicks:

he ranks david lee one of the top 15 power forwards in the league:
http://www.wagesofwins.com/15PF0708.html

every season he has ranked eddy curry as one of the top 15 overpaid players in the league.
http://www.wagesofwins.com/Over41.html
http://www.wagesofwins.com/EddyCurry.html

jamal crawford last year was ranked one of the top 15 most overrated players in the league
http://www.wagesofwins.com/OverRatedPPG0708.html

how he ranked our team as a whole last year
http://www.wagesofwins.com/Knicks0708-76.html
Hint: our above average performers in '07 were lee, randolph and balkman. Most of our team are below average players and more than a handful have NEGATIVE value while they are on the court.

I would agree with most of those findings.

Almost everybody agrees that the margin a team beats other teams by, controlled for by possessions per game, is by far the best predictor of future winning percentage for a given team. When you sum up the player win scores on a team, you are inadvertently summing up to the margin a team beats other teams by, controlled for possessions per game.

As fun as run and gun is to watch, I still firmly believe that you win games in the postseason by your defense, and controlling the number of possessions each team gets (to the extent possible). Which is why I've always thought Phoenix was just eye candy, a regular-season wonder. But I'm willing to see where 'Antoni can take us, and I'll support him early. I wanna' see progress though...

Berri has actually been pretty positive about the pheonix team. Though pheonix lets teams score a lot of points, with the exception of last season, Berri would probably argue that pheonix was actually an ABOVE average defense. When you control for the number of points pheonix allows per possesion, pheonix actually was actually a fairly stingy defense, when they dropped to the middle of the pack. One could argue that Dantoni installed a much better defensive system than he is generally given credit for, which gives reason for us knicks fans to hope.

On the down side, Berri also argues that coaching changes don't have considerable impact on player performance. Berri argues that in the aggregate, the main way coaches tend to impact a team's winning percentage is by changing how they allocate minutes to various players.

On the plus side again, On his Blog Berri has derided Isiah for his coaching decisions. So while coaches can't hugely improve their team's winning percentages (other than by giving more minutes to the best players), Berri argues that it is quite possible that Isiah through his incompetence, managed to hold the knicks back.

But overall, no matter what D'antoni does, the knicks are in for a rough season. As of now the only above average players we have are David Lee (with a Wins per 48 minutes of over .200 if David Lee played every minute of every game, it would translate into about 17 wins for the season) and Zach Randolph (rated slightly above average). So unless some of our younger players improve, our ceiling is probably around 30 wins.