No correlation between salary and wins?

[EvanZ]

As the NBA league office continues the lockout in an apparent bid to create more parity among teams, a professor of economics at Smith College who has studied the issue says there is almost no relationship between the size of a team's payroll and its success.

“The statistical correlation between payroll and win percentage is practically nonexistent,” said Andrew Zimbalist.



Read more: http://basketball.realgm.com/wiretap/21 ... z1ab1qSdtH

"Practically nonexistent"...ok...

Code: Select all

> summary(salary.lm)

Call:
lm(formula = PYWINS ~ Salary, data = salary_wins)

Residuals:
    Min      1Q  Median      3Q     Max 
-24.186  -7.223  -1.698   6.755  25.239 

Coefficients:
            Estimate Std. Error t value Pr(>|t|)   
(Intercept)  14.1752     7.9204   1.790  0.08432 . 
Salary        0.5047     0.1433   3.523  0.00149 **
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 

Residual standard error: 10.8 on 28 degrees of freedom
Multiple R-squared: 0.3071,	Adjusted R-squared: 0.2824 
F-statistic: 12.41 on 1 and 28 DF,  p-value: 0.001485 

Almost a 30% explanation of PyWins by NBA payrolls is "practically nonexistent"?

[Crow]

I'll probably chalk that up as an expert simplifying things talking to a reporter and a lay audience.

Zimbalist has an extensive academic and consulting record in sports economics.


Still...

In the recent Win Shares thread I noted that the last 5 champs averaged being $22 million over the cap (and that is about twice as far over the recent caps as the median spending teams). I just checked the previous 7 years and it was an average of about $11 million over the cap. So the most successful teams in the last 5 years doubled the additional spending over the cap compared to the average of the previous 7 champions. But apparently they didn't know what they doing or didn't need to or didn't gain much from it?

Spending league wide on its own with no other criteria fulfilled doesn't ensure success but other studies could be constructed (multi-variate regression?) that might show a significant plus for spending when other key ingredients- playing, coaching and management talent- are above average (with perhaps the playing talent being judged- in some fashion- before and after the spending above the cap or league average). I would think. Probably ought to find or re-read the previous academic articles on the subject before I say much more.


A quick blog post find, though:

"One key observation… in the last six years if you were willing to spend over $400 million your team won at least 250 games and a trip to the conference finals, except for New York."

http://wagesofwins.net/2011/09/18/how-n ... eir-money/

So 8 of 9 teams who spent over $400 million (it looks like only slightly above average spending for the 6 year period) got at least one final 4 appearance. And in fact they got 15 of the 24 such appearances or more than double the number expected for 9 teams if they didn't have some edge... from talent... and spending on talent.

And another:

"Payroll does not explain much of wins in the NBA, MLB, or NFL. Specifically, payroll only explains 12% of the variation in wins in the NBA."

http://wagesofwins.net/2006/05/24/summa ... n-the-nba/

So I ask is there something different in the methodology or is there a trend over time toward salary explaining more of wins generically? I'd like to see more about that given that Evan reports 30% of wins explained by spending this past season compared to the 12% Dave Berri reported 5 years ago. There might not be a stable trend, it could bounce around, but I'd still like to know the trend in addition to a couple point estimates of scale of significance.

What is the record of teams with significant "advanced analytic" efforts vs the rest?

How do the correlations compare across sports? Are NBA general managers doing better than other sport GMs (baseball was reported by Berri to have 24% correlation of spending and wins in 2006) and is it because of their managerial and analytic talent or the nature of the game?


Again, I'd ultimately place more importance on looking at impact of spending on conference finalists or title winners than regular season wins.

[EvanZ]

In the first place, it seems like a strange hypothesis, anyway. If there were no correlation between salary and wins, would that be so surprising in a league with a salary cap? The fact that there is such a large correlation is primarily due to the allowance of players to be signed to huge amounts if owners are willing to pay a luxury tax. And bad teams lacking stars that deserve max contracts.

Take away the salary cap completely, and is there any doubt that the correlation between salary and wins wouldn't approach 1? Cuban probably would've signed LeBron and Wade last season if he could have.

[Crow]

http://www.jeremyscheff.com/2011/07/the ... d-by-team/

"Another apparent trend is that, in the past few seasons, the correlation between wins and payroll is much higher than it was previously."

i think this work is xkonk's.

[Crow]

Phil Birnbaum's take:

http://sabermetricresearch.blogspot.com ... ation.html



More discussion:

http://sportskeptic.wordpress.com/2011/ ... -a-twofer/

[xkonk]

http://www.jeremyscheff.com/2011/07/the ... d-by-team/

"Another apparent trend is that, in the past few seasons, the correlation between wins and payroll is much higher than it was previously."

i think this work is xkonk's.

Nope, not mine. Although I have to say with the salary data I have, I get different correlations than Jeremy reports (for example, in 2009-2010 I get a correlation of .465, not .57. My data is from HoopsHype. If you look from 2002 to 2010 (previous years from USA Today), the correlation is only .256. If you run it as a regression and add a salary squared termed, you find a negative coefficient for the square - diminishing returns of wins from added salary. Adding season as a term doesn't result in any significant effects or interactions, so any potential changes in correlation over years is likely small at best.

Either way, I think the argument in the Sport Skeptic post is reasonable. Salary is a stand-in for talent, but an imperfect one. So yes, you would expect that teams that spend more would win more, but it isn't a guarantee by any stretch. If you were to account for team quality as Crow suggests, I would be shocked if salary adds any explanatory power.

In regards to Evan's hypothetical of taking away the salary cap and the correlation moving to 1 - baseball has only the softest of caps, with the only restraint being revenue sharing. Dave Berri's article (and book) note that the salary-wins correlation in baseball is still not that high. Berri's blog has also reported research showing that adding caps or making caps 'harder' has little to no effect on parity, which I think also speaks against that idea.

In one of his recent posts Henry Abbott at TrueHoop suggested that the real benefit to spending a lot is that you can make a mistake. Atlanta, for example, doesn't spend a ton and so they're hamstrung when they sign Joe Johnson to a huge contract. Dallas or LA could sign Johnson but still be willing to spend even more and bring in another player, presumably one actually worth the money (Abbott used the example of Dallas signing Haywood and then still getting Chandler). That second player mitigates the poor play of the first, particularly if the team is willing to play the first guy less despite his contract. If true, it still doesn't speak well of GMs' ability to pick talent; they should just get the good player the first time and not pay extra to make up for an initial mistake.

[Crow]

xkonk,

Sorry for my confusion, I thought I saw a connection of your name to that website but I guess I was mistaken.

[Crow]

More discussion of the quote and the issue:

http://forums.realgm.com/boards/viewtop ... &t=1134016

[Mike G]

Would a low correlation for much of the past decade be attributed to one or two spectacularly bad money managers with wealthy teams?

In 2006, the Knicks had on their payroll : Houston, Marbury, Hardaway, Davis, Taylor, Anderson, Curry, QRich, JWilliams, Crawford, Rose, and James, all making at least 5 million that year.
According to Win Shares, this bunch totaled 12.1 wins contributed to NY, at a cost of $121 million.
http://www.basketball-reference.com/teams/NYK/2006.html

At $10 mill per win, that brings down the efficiency of the whole league, such that a legion of penny-pinching GM's would struggle to offset it and make the league look like wise spenders. In terms of $/W, that is.

Supposing $1,000,000 per win is par for the course, if 29 pretty good GM's manage to get wins for an avg of $900,000 each, their 1207 league-wide wins cost $1.09 billion.

Adding in the Knicks' 23 wins and $0.13 billion, the league averages $990,000 per win: On average, 30 mediocre GM's.

[Mike G]

In one of his recent posts Henry Abbott at TrueHoop suggested that the real benefit to spending a lot is that you can make a mistake. Atlanta, for example, doesn't spend a ton and so they're hamstrung when they sign Joe Johnson to a huge contract. Dallas or LA could sign Johnson but still be willing to spend even more and bring in another player, presumably one actually worth the money ...

Isn't this actually "the real benefit to having a lot of money ..."
It doesn't benefit you to just "spend a lot", if you don't have that kind of money.

[xkonk]

In one of his recent posts Henry Abbott at TrueHoop suggested that the real benefit to spending a lot is that you can make a mistake. Atlanta, for example, doesn't spend a ton and so they're hamstrung when they sign Joe Johnson to a huge contract. Dallas or LA could sign Johnson but still be willing to spend even more and bring in another player, presumably one actually worth the money ...

Isn't this actually "the real benefit to having a lot of money ..."
It doesn't benefit you to just "spend a lot", if you don't have that kind of money.

No, I mean spend a lot. Obviously you can't spend what you don't have, but I assume that potential owners wouldn't be allowed to buy an NBA team if they weren't willing and able to cover at least a certain amount of expenses. And there are owners with a lot of money who don't spend it on their team; in 2005 there were apparently 13 owners representing 12 franchises who were personally worth over $500 million each (http://www.usatoday.com/sports/basketba ... owners.htm). But many of them aren't what I would call free-spending teams (e.g. Detroit, Indiana, the Clippers). Guys like Cuban might have an advantage by having deeper pockets than other owners, but his real advantage is a willingness to spend it.

[Mike G]

OK, so Cuban is like a poker player who doesn't like to fold, who would rather bluff and sometimes get burned, while as often as not out-bluffing the rest of the table.

Another owner, having bet big and lost, becomes gun shy and can't abide another big loss, so later gets out of the game quickly.

It's partly deep pockets, and it's partly willingness to gamble. Some owners didn't get into it for the gambling aspect, and they're probably appalled that it's part of the game. Their advisers goad them into showing they mean business, and against their better judgement they toss in a tenth of a billion; then they're shocked by how little they get from an investment.

In an unregulated poker game, the richest players will always win the pot: they just bet such sums that others cannot cover them and have to fold.

Maybe if there were no salary cap, ever-richer gamblers would buy NBA teams and outbid the weaker owners to get all the best players. Disparity would increase, and the finals would always be between the richest teams. Even players for poor teams would make more money. But their fans may not like it.

[Crow]

I wrote Andrew Zimbalist advising him of the discussions here and at Real GM and asking if he had any comment. He replied and asked me to post this comment:

"if one runs team payroll on team win pct in a given year, the correlation coefficient is rarely significant in the NBA (also true for the NFL). actually, Howard Beck lopped off part of my quote. I did say that in a few years there was a significant correlation at the .10 level."

[Bobbofitos]

In an unregulated poker game, the richest players will always win the pot: they just bet such sums that others cannot cover them and have to fold.
.

Poker hasn't been played like this for the past 30 years or so

[Mike G]

Exactly. And as far as I know, nobody is lobbying for elimination of the salary cap.