Wednesday, February 6, 2008

Wondering about the Wages of Wins

3 Shades of Blue in general (and myself in particular) have always been fans of the work of David Berri, one of the co-authors of the book Wages of Wins. I recently had the opportunity to ask David some questions via email about the Grizzlies and the Pau Gasol trade in particular. Here are the results of that conversation.

3SoB: David – First let me say how much I appreciate you taking the time to do this. I suppose given your Bakersfield residence you felt obligated the fill in the blanks left from the Pau Gasol trade. After all if Memphis can give the Lakers a championship you can at least give a bit of your time to the Grizzlies blog writers!
David: Glad to talk to you and your audience (which I gather is growing very rapidly).

3SoB: You wrote a book called the ‘Wages of Wins’ a few years back in which you attempt to use statistics to determine a player’s value and how many wins he actually produces for his team. Would you give us a Cliff Notes version of how this works?
David: Here is the overview of the Wins Produced model:
The model starts with the basic idea that wins in the NBA are determined by offensive and defensive efficiency (points divided by possessions). From this relationship, one can estimate the value of points, rebounds, steals, turnovers, personal fouls, field goal attempts, and free throw attempts. Additional statistical work gives us the value of a blocked shot and an assist.
Additional notes
1. The value of each statistic is derived from an econometric (or statistical) model. And the values are an estimate of the impact the statistic has on the outcomes we observe for the team.
2. The analysis indicates that points, rebounds, steals, turnovers, and field goal attempts have virtually the same impact (in absolute terms) on wins. Personal fouls, free throw attempts, blocked shots, and assists are worth somewhat less than the other factors.
3. When we look at the impact of these statistics we learn two lessons:
a. In terms of scoring… efficiency matters. If you score points inefficiently you do not actually help your team win.
b. Possession variables – rebounds, turnovers, steals – impact outcomes. Players that don’t score, but get you rebounds and steals (and/or avoid turnovers), can have a significant impact on wins.
4. One last point… it is assumed that a player statistics are his statistics. In other words, if you score 10 points (or get 10 rebounds), you get credit for these points (or rebounds). If you take 5 shots, you are charged the cost of these shots. It’s important to note that there is some interaction between the player statistics, but I have found this effect to be rather small. For the most part, what you see in the NBA is what you get. So it makes sense to credit a player for his statistics (by the way, I would not make the same argument for NFL players)

3SoB: You wrote on your website that this trade makes the Lakers the favorites to win the West of they can get Bynum back. You went so far as to say that Gasol’s presence puts the Lakers among the elite in the West even without Bynum. From someone who remembers the Grizzlies 0-12 playoff record with Gasol, is this actually a playoff prediction you are making or just a regular season record assumption?

David: Let me start by reviewing the playoff history of Gasol on Memphis.
In 2005-06 the Grizzlies were the 5th best team (in terms of efficiency differential) in the NBA. But they faced Dallas in the first round, who happened to be the 3rd best team in the Association.
In 2004-05 the Grizzlies were the 7th best, facing the second best (Phoenix).
And in 2003-04 the Grizzlies were the 9th best, but faced the very best team in the league (San Antonio).
In sum, Gasol and Memphis were never favored in the first round. Yes, it is not beyond reason that they could have won a playoff game. But it was not expected that Memphis was going to win any of these playoff series. So judging Gasol on the Grizzlies lack of playoff success is a bit unfair.

Now about the Lakers…
Assuming Andrew Bynum and Trevor Ariza are healthy (a big if), here is what Gasol does for this team.
Gasol not only takes the minutes of Kwame Brown, but also probably some of the minutes of Vladimir Radmanovic and/or Ronny Turiaf. Gasol is more productive than all three, so that’s a big help.

In addition, his presence allows the Lakers to play Lamar Odom more minutes at small forward, where his value will be higher. Finally, and this is a smaller positive, Javaris Crittenton was easily the least productive guard on LA’s roster. So Crittenton’s departure helps a bit.

It’s important to note that the Lakers ability to challenge Boston depends on the health of Bynum. If Bynum is not healthy, then the Lakers do not look much different from the other elite teams in the West.

3SoB: What does this leave the Grizzlies with this year, according to your math? Is it possible the Grizzlies can surprise this year with unproven young players getting a real chance to perform?
David: There is a great saying from Mike Tyson…'Everybody has a plan until they get punched in the face.' (I took this from a Peter King column, and King was quoting Michael Strahan).

When I read this quote I was reminded of everyone who fought Tyson when he was in his prime. These fighters would tell themselves if they tried hard and believed; they could beat Tyson. Each of these fighters would enter the ring with a positive attitude and a clear plan to achieve victory. And then Tyson would hit them in the face and the plan would go to hell.

I tell this story because the Grizzlies are now in an Association where most every other team is Tyson. And the Grizzlies are telling themselves that if they try really hard, and execute their plan, they can shock the world.

Unfortunately, soon after the ball is thrown in the air at the start of each game, reality is going to set in. Memphis has too many holes to compete successfully most nights.

Look at the center and power forward positions. With Gasol gone, Kwame Brown becomes the most productive player Memphis employs at the four and five spot. And Brown is well below average. The other centers are Jason Collins and Darko Milicic, two players whose production of wins this year is in the negative range. At power forward the team is forced to play Hakim Warrick and Rudy Gay, two players who do not rebound well enough to play the four spot. At the other three spots on the floor there is some production. Mike Miller is one of the most productive shooting guard-small forwards in the league. And Mike Conley and Kyle Lowry are productive point guards.

But right now, those three players are the only above average performers on the team. Remember, wins in the NBA are not magical. Wins happen because your players are productive. Or to put it another way, if your players are not productive, you don’t get to win very much.

Given what your players have done this season, it would be truly surprising if Memphis was able to win games consistently the rest of the way.

All that being said, Memphis may win 10 more games the rest of the season. So 10 times you will get to say “if only they played like this every night, they would be pretty good.” But that line of reasoning is incorrect. Bad teams will win games and good teams will lose once in awhile. It’s what a team does over time that is the mark of how good or bad a team is, and I think over time – given this roster – the Grizzlies are going to disappoint their fans.

3SoB: Memphis is definitely suffering this season but I seem to remember something you wrote that implied a player’s best performance years don’t begin until their 3rd season in the league. That means Lowry, Conley, Gay, Hakim and even Darko if you eliminate his wasted two years in Detroit (he is only 22 after all) lie ahead of them. Throwing in some decent drafts and the team can progress in the future can’t it or am I just wishing on stars now?

David: If the players improved and/or the team adds more productive players, then of course the team improves.

Let me talk about improvement first. It is the case that young players can get better. But I think people go too far with this tendency. Basically people tend to see all young players as “future stars.”
Here is some data that might damper your enthusiasm. Looking back at the drafts from 1991-1995, here are the number of players in each draft to play at least 10,000 minutes in his career and post a career WP48 higher than 0.150 (average WP48 is 0.100 and I think 0.150 is a mark of someone who is “good”).
1991 – 4
1992 – 5
1993 – 3
1994 – 6
1995 – 2
There were 93 players drafted from 1991-95 who played 10,000 minutes in their NBA career. Of these, only 20 were significantly above average players throughout their careers. In other words, only 22% of players who played significant minutes in their career managed to be significantly above average players.
When people look at young players like Lowry, Conley, Gay, Warrick, and Milicic, they tend to think that each of these players could become “good.” But the data tells us that most young players never become “good.”
The early returns on these players do suggest that Conley and Lowry are going to be above average players. And I think Gay can be above average if he plays small forward. But I am not sure that Warrick and Milicic are going to develop into players who make a significant contribution to team success.

As for picking better players, that’s always possible. Let me talk about that below.

3SoB: Along those lines how long do most NBA players maintain their level of performance? Is there a general rule that says after a certain age most players start to decline in performance?

David: In general players get better when they are young and worse when they are old. But I don’t know the parameters of the curve (yet). This is one topic I wish to look at in the future.

3SoB: Chris Wallace has a lot of cap space available to him now. What upcoming free agents do you believe produce the most wins and who produces the least? (You can use the list provided at for a guide here). What about college players?
David: Here are the top 20 unrestricted free agents (among players who played at least 500 minutes in the first half of 2007-08). The players are ranked in terms of WP48 [Wins Produced per 48 minutes] over the first half of this season.

Player / Minutes Played / WP48 / Wins Produced
Shawn Marion (ETO) 1,473 / 0.314 / 9.6
Kurt Thomas 794 / 0.302 / 5.0
Baron Davis (ETO) 1,601 / 0.237 / 7.9
James Jones (P) 674 / 0.212 / 3.0
Brent Barry 541 / 0.207 / 2.3
DeSagana Diop 740 / 0.177 / 2.7
Antawn Jamison 1,612 / 0.175 / 5.9
Eddie House 814 / 0.163 / 2.8
Corey Maggette (ETO) 1,236 / 0.157 / 4.0
Anthony Carter 860 / 0.140 / 2.5
Grant Hill (P) 1,154 / 0.133 / 3.2
Bonzi Wells 907 / 0.132 / 2.5
Allen Iverson (ETO) 1,710 / 0.128 / 4.5
James Posey (P) 791 / 0.126 / 2.1
Carlos Arroyo 616 / 0.121 / 1.6
Sam Cassell 755 / 0.119 / 1.9
Anthony Johnson 914 / 0.119 / 2.3
Mickael Pietrus 637 / 0.098 / 1.3
Beno Udrih 1,188 / 0.091 / 2.3

Here are the top 15 restricted free agents (among players who played at least 500 minutes in the first half of 2007-08). The players are ranked in terms of WP48 [Wins Produced per 48 minutes] over the first half of this season.

Player / Minutes Played / WP48 Wins / Produced
Jose Calderon 1,277 / 0.319 / 8.5
Andris Biedrins 1,130 / 0.303 / 7.1
Josh Childress 1,068 / 0.219 / 4.9
Emeka Okafor 1,359 / 0.203 / 5.7
Dorell Wright 649 / 0.170 / 2.3
Carlos Delfino 1,015 / 0.170 / 3.6
Andre Iguodala 1,607 / 0.165 / 5.5
Luol Deng 1,145 / 0.152 / 3.6
Ryan Gomes 1,128 / 0.131 / 3.1
Paul Millsap (T) 862 / 0.130 / 2.3
Monta Ellis 1,431 / 0.122 / 3.6
Josh Smith 1,426 / 0.114 / 3.4
Daniel Gibson 1,274 / 0.073 / 1.9
Kelenna Azubuike (P) 987 / 0.070 / 1.4
Ronny Turiaf 637 / 0.059 / 0.8

As you can see, there are a few free agents who will help. DeSagana Diop is an interesting possibility, since he has been above average before this season. James Jones is a big reason the Blazers have improved, but he has not been productive in the past. So I am not sure he will continue to perform like this in the future.

As for the draft, Erich Doerr posted a study at the Wages of Wins Journal a few days ago. One name that leaped out was Kevin Love.

Love would give the Grizzlies what they need at power forward. So he would help. Of course he has to declare for the draft. And the Grizzlies have to select him. And he has to produce as a pro. So there are quite a few ifs with respect to Love and the draft.

I think you should see, though, that it is possible for the Grizzlies to get better in the future. But a few things I would emphasize:
1. I am not optimistic about this team as it is currently constructed.
2. It is possible for this team to improve, but it is going to need to get more productive players.
3. I do not think the Gasol trade is going to help the Grizzlies get more productive players. At least, I am not optimistic about trying to improve a team via low first round draft choices.

3SoB: Any other comments you would like to make?
David: Thanks for the opportunity to answer your questions. I wish I could give fans of Memphis more hope in the short-run.

David Berri is a professor at CSU-Bakerfield and writes the Wages of Wins Journal where he is always putting up timely and informative blogs. We at 3 Shades of Blue highly recommend you checking out his site.

BallHype: hype it up!


Tim said...

Good interview. I get this guys points but I think he is too based on statistics and numbers. I base way more on heart and desire of players. Where are his stats of how many times Pau checks for blood or cries about calls and his body language that effects the whole team when he is having a bad night. Or how many post players have the career nights on Pau (Horford had 20 rebounds on him last night).

And im not saying I dont think the Lakers wont be good with Pau but theres more than just his stats and winning thing.

Because what if next year Rudy totally takes over the role as leader and brings the team to new heights. I know it doesnt happen like that a lot but it does happen. Gilbert Arenas was i guy that didnt look like a franchise player early on and he shocked everyone.

Anonymous said...

A book that contains a number of flawed analyses. Some people are easily dazzled by stats, even when they are clearly being manipulated to produce a desired outcome.

David Jones said...

Anon - All David Berri is saying in this post is that the Grizzlies are terrible and will be for a while. I say that's a pretty fair assertion.

As for his metrics, please be more specific about why Berri's reasoning is flawed. Otherwise, your comment is uninteresting. It seems just as likely that Berri slept with your wife as you have some interesting criticism of Berri's work.

If not, let's hear it.

merl said...

The chief criticism of WoW is that it overvalues rebounding.

Berri's contention is that a player's rebounding totals are relatively static from year to year, implying that it is the player who is responsible for all of the value of their rebound.

This ignores several points:
1. that rebounds may be partially based on the role a player is asked to perform on a team (partially offset by the mystical position adjustment in WoW)
2. The diminishing returns associated with rebounding
3. Numerous anecdotal studies that show that despite assertions to the contrary, superior rebounders *do* appear to be grabbing rebounds that their teammates would otherwise get, otherwise a superior player rebounding who got say, +500 rebounds per year more than the average player at their position would be playing on a team that was also +500 rebounds vs average rebounding team. Unfortunately when players pull down a prodigious number of rebounds this never seems to carry accross to the team level. Although Berri assumes that great rebounders don't take rebounders from their teammates, for some mysterious reason great rebounders always seem to play on teams with poor rebounders. The end result is that the WoW metric gives the above average rebounders a disproportionate amount of credit.

The other factor is that the point at which Berri considers shooting to be 'efficient' and thus worth rewarding is at 50%, which means that a player who shoots 15/30 every night receives the same score as a player who shoots 1/2 every night. I don't regard this as as much of a problem as the rebounding valuation however, since the break even point of PER is 30%, which I think is more incorrect than WoW with regards to shooting percentage . Since every player in the league shoots > 30%, PER basically rewards high volume shooters, I guess working on the notion that if a coach lets a player gun away, that must be a good player.

merl said...

In addition, defensive rebounds (that have a value of 1) are a statistic that captures all of the defense that caused the missed shot, as well as the rebound.

So giving a score of 1 to the rebounder is crediting them with the rebound, but also crediting the rebounder with ALL of the defense that was played on that play.

Anonymous said...

Basically, he makes you think that his model has excellent predictive power. But it's not really 'his' model that has excellent predictive power, but what his model will eventually generate - points-differentials (or offensive, defensive efficiencies, as people like to call it).

People have already known that points-differential for a team is highly correlated with the number of the wins that a team has. He just found a relationship between points differentials and individual categories (rebounds, points, etc.).

However, there in lies the fundamental problem of his model. There're INFINITELY many ways of assigning weights to different categories. He makes you think that there's only one way of assigning weights - he arbitrarily gives you equations for computing possessions. But it's really easy to come up with another equation that makes perfect sense but will give you different numbers for individual players. With some 'adjustments' (that he makes a lot of) these different weights will still all add up to correct points differentials, thus will have the exact same predictive power.

Additionally there are a lot of 'adjustments' that he makes that seem very arbitrary.

David Jones said...


So if I understand correctly (and I haven't read the book), Berri says x individual statistics are most important for determining the outcome of the game.

However, he favors certain individual statistics, like rebounding, and discounts others which may be just as likely to produce or predict wins.

So the question for the Grizzlies is, what individuals can they add that will produce more wins.

It's clear that adding a seven foot Spaniard that scores at will but doesn't play defense or rebound agressively won't produce wins for the Grizzlies.

So nor that we have financial flexibility, what 'individual characteristics' will produce wins?

Costa said...

It's important, when reading statistical models to both be guarded, but also be open-minded.

As such, Dave Berri's WoW numbers shouldn't be accepted immediately as gospel, but they shouldn't be ignored either. Does he overvalue rebounds? Perhaps. Does he underrate a player's ability to create his shot? Perhaps. But maybe there's some truth there too. Maybe the average person, who's so in love with scoring and PPG numbers, doesn't realize how much other things like rebounding or shooting efficiency figures in.

As observers, it's important for us to use the correct judgment and logic when looking at these numbers, but the fact that Mr. Berri gives us another set of data to combine with other ones at our disposal such as the subjective stuff that Tim talks about in the first comment, Points-per-game, Rebounds-per-40-minutes, Player Efficiency Ratio, Field-Goal-%, Assist-to-turnovers and Pace can be very valuable.

Observing all these factors with an open mind and then matching them up with what we see with our own eyes while watching the games is the best way for us to get the deepest understanding of the game as possible. While I have my reservations with some of his logic, I'm happy Mr. Berri has given us this tool. I feel it has opened my eyes to some aspects of the game I would otherwise have missed. =)