Wednesday, May 14, 2008

Baseball Salaries still not correlated with performance

There are lies, there are damned lies, there are statistics, and then there are baseball statistics.

A number of hits from TheBigLead sports blog to an earlier discussion of baseball salaries vs. team performance has prompted me to check that conclusion again with more data. The conclusion was that there is little correlation between salary and winning games. Later, Paul suggested we read Moneyball, I read The Baseball Economist instead.

The Big Lead helpfully pointed to the USA Today baseball salaries database where I collected total team salaries going back to 1988. I also collected the wins for every team through those years (2007 to 1988). So now I have more data to support the boring conclusion that team salaries don't seem to correlate with team wins.

In order to eliminate the change in salaries from year to year I normalized the yearly salaries of each team to the median (half the team salaries above, half below) salary for that year. The median salary is set at 100%. I also normalized the team wins for each year to the median number of wins (usually about 81, since 162 games are played and over the entire set of teams half are lost and half are won). This normalization makes sense because it costs some amount to run a baseball team, but we are interested in seeing if paying more than that amount, or more than the other guys increases our ability to win more games than the other guys and make it to the playoffs.

The plot (click it for larger) shows that the effect is weak, an increase of 10% over the median team salary yields only a 1.7% increase in games won over the median (about 1.4 games) and the correlation only explains about 15% of the difference in winning. 85% of the difference are due to other factors. This is true with and without the traditional Yankees team salary outliers from 2003 through 2007.

What if we look at last year's salary's effect on this year's performance? We get an even smaller effect and a worse correlation.

So if salaries don't have a real effect on baseball wins and wins is what managers and owners aim for, why do they keep paying the big salaries? Paul says read Moneyball, and some teams are using SABRmetrics and other statistics to find the biggest bang for their player buck, why isn't it reflected in the salaries vs. wins? Nearer and dearer to my heart, why did the Detroit Tigers choose to pay a $137 million, team payroll (#3 below) and they are in last place in the central division, behind Kansas City for goodness sake?

This year's salaries:

rank team salary
1 New York Yankees $209,081,577
2 New York Mets $137,793,376
3 Detroit Tigers $137,685,196
4 Boston Red Sox $133,390,035
5 Chicago White Sox $121,189,332
6 Los Angeles Angels $119,216,333
7 Los Angeles Dodgers $118,588,536
8 Chicago Cubs $118,345,833
9 Seattle Mariners $117,666,482
10 Atlanta Braves $102,365,683
11 St. Louis Cardinals $99,624,449
12 Philadelphia Phillies $98,269,880
13 Toronto Blue Jays $97,793,900
14 Houston Astros $88,930,414
15 Milwaukee Brewers $80,937,499
16 Cleveland Indians $78,970,066
17 San Francisco Giants $76,594,500
18 Cincinnati Reds $74,117,695
19 San Diego Padres $73,677,616
20 Colorado Rockies $68,655,500
21 Texas Rangers $67,712,326
22 Baltimore Orioles $67,196,246
23 Arizona Diamondbacks $66,202,712
24 Kansas City Royals $58,245,500
25 Minnesota Twins $56,932,766
26 Washington Nationals $54,961,000
27 Pittsburgh Pirates $48,689,783
28 Oakland Athletics $47,967,126
29 Tampa Bay Rays $43,820,597
30 Florida Marlins $21,811,500

No comments: