Simple vs. Complex Offenses, Part 3

In the last post, we came up with a quantified measure of the complexity of an offense based on the range of its called formations. Now, we’ll take those complexity numbers and compare them to actual offensive output to see if there’s a relationship to try and answer the million dollar question: what's better, simple or complex?

There are probably more detailed approaches to answer this question, but a good starting point is just comparing complexity scores to an objective offensive measure and seeing if there's a relationship. Below is a scatter plot of each of the charted teams’ offensive complexity scores to their offensive F/+ rating for the 2013 season:

It’s by no means a perfect fit, but an R-squared of 0.2 definitely points to at least a little bit of a relationship. And the trendline is sloping downwards, indicating the more complex an offense, the worse it does on average. (Kansas fans, I hope you find this a little validating.)

Now, the standard disclaimer: correlation doesn’t imply causation. Maybe offenses increase their complexity because they have below-average players and they’re constantly trying to find the right set of play calls. That argument gives me a little bit of pause, though. In theory, programs should have their identity set from the start and recruit players that fit that identity accordingly. If you’re constantly changing the core of your offense to see what works, that also doesn’t provide a lot of stability in the middle of a season. It’s hard enough to get an offense right in training camp, let alone a constantly changing one in the middle of an 11-game season with classes in full swing. It’s reasonable to speculate that sticking with your plan probably provides a better hope of success for below-average players than mixing things up, even in the face of enormous pressure to make changes if the plan isn’t working.

This is by no means a complete data set. It’s one season with fewer than 30 teams, hardly enough to draw sound conclusions. But thanks to charting data, at least now we have the information, and the approach, to answer these types of questions with actual analysis. The answers will get more conclusive with additional data, so by all means, go sign up and help the cause!