First, I do not mean to steal a post idea from Jimmy Kempski with this topic. He inspired this post with the tweet below, and I suppose there’s a good chance that tweet was essentially a preview of an upcoming post of his. But I couldn’t wait for him to do the research, so I had to explore myself:
WR and OL is taking up 41.7% of the #Eagles' salary cap. The entire defense takes up 36.9%.—
Jimmy Kempski (@JimmyKempski) March 01, 2014
- Where do the Eagles rank in percentage of spending on offense vs defense?
- Is there an ideal ratio for spending on offense vs defense?
I was especially curious about this because I recall reading an Andrew Brandt article from 2011 on analytics in football, in which he stated:
One example of our analysis showed the most successful teams were balanced in their spending on offense and defense, a trend that stayed true most years. Teams with large disparities in spending on one side of the ball tended to have less sustained success. That is information I shared with the scouts and we used in a way that could guide us.
This was just one illustration; there were many more that will not be shared here.
Brandt is best known for his time as Vice President of the Green Bay Packers from 1999-2008, but he also consulted for the Philadelphia Eagles in the 2009 offseason. In a role that, per his Wharton bio, he actually focused on Salary Cap management. If Brandt was willing to share this piece of information to the Internet, it’s almost certainly something that he would have preached to Eagles executives like Howie Roseman, a cap management aficionado himself. So shouldn’t we expect the Eagles to have relatively balanced spending?
Kempski’s tweet makes it look like the Eagles must be overspending on offense and underspending on defense, but context is needed. How does the Eagles’ ratio compare to what other teams had for offense vs defense cap spending in 2013?
Using overthecap.com’s 2013 estimates, I determined each team’s percentage of cap spent on offense and defense (so this ignores special teams), and the Eagles came out as the #1 most imbalanced team. This is a surprise given Brandt’s Eagles connection.
The league average was 52-48 offense-defense. Here’s the chart.
Now, is that bad? Does imbalanced spending lead to poor performance?
Brandt’s statement addresses “sustained success.” I couldn’t find data for years prior to 2013, so we can’t look for any year-to-year trends, but we can still look at whether balanced teams did better or worse this past year.
I measured imbalance as how many percentage points both offense and defense are away from 50%. The Eagles, at 66-34, had an imbalance value of 16. The Bengals, at 39-61, were given an 11.
To measure success, I used teams’ 2013 Pythagenpat records, as calculated by Chase Stuart at Football Perspective. These scores are slightly better indicators of how well teams played than win-loss records. (Read Chase’s article for an explanation.)
Then I plotted imbalance vs success:
So based off of just one year of data, it really doesn’t look like there’s much of a relationship between balanced spending and on-field success.
That concludes my exploration. But I failed to demonstrate any evidence for any particular optimal spending ratio. Does anyone have any ideas for what they think it should be? Two things to keep in mind:
1) Several sources have established that offense is about 1.5 times as important as defense at producing wins.
2) On the other hand, Chris Brown of SmartFootball has said that studies have shown that scheme may be able to mitigate the value of offensive talent more than it can for defensive talent. I have not found these studies, but Chris is highly credible, and the idea matches our intuition. If you can find these studies, please, please comment.
 Is it even possible to steal a post from someone if you have 30 followers and they have thousands?
 Caveat: Brandt’s statement does not make it clear whether he means raw spending or spending against the cap. I only analyzed his hypothesis with respect to cap spending because that’s all I could find. The numbers are naturally highly correlated anyway, but it is a potential flaw of this analysis.