The traditional football stats, as we've come to understand them -- total offense, total defense, third-down conversion, etc. -- seem too often to miss the point of statistics, which is help us understand why one team won over the other.
Inspired by work in recent years at Cold, Hard Football Facts, we have been working with a statistician over the last several months to crunch some numbers in an effort we hoped would help us find some answers.
What CHFF did was to look at professional football statistics and try to understand which ones correlate to winning. While on that path they experimented with new statistical formulations, such as "scoreability" and "bendability," all in an effort to help us understand which stats have significance and which ones don't. They call them "Quality Stats."
Working with college stats poses some different challenges, as opposed to professional stats, among which are:
- There are 32 professional teams that operate under a set of rules that are designed to insure some degree of parity among the teams. There are 120 FBS division teams, and while there are some rules in place designed to help insure competitiveness, those levelers aren't as prevalent as they are in the NFL.
- NFL teams play a 16-game regular season schedule against teams all within its same level of competition. The FBS teams play a 12-game regular season schedule, and many play at least one game against an FCS opponent – teams that operate with far less resources in terms of scholarships it can grant, facilities, and perhaps most importantly, television exposure.
- The Bowl Championship Series has created a group of 64 teams that have competitive advantages over the remainder of their FBS brethen, to the extent that not being a member of the BCS is seen as a decided disadvantage when it comes to recruiting.
In an attempt to decrease the skewing that can take place in the stats because of the wide variances of competition within college football, we've looked at only BCS games – games between teams in BCS conferences.
Of course there is a difference in the competitive levels even within the BCS, and obviously the numbers of some extremely competitive teams like Boise State and TCU get left out of the stats. In addition, we're all familiar with the major FCS upsets over FBS teams in recent years. However, you've got to draw the line somewhere unless you want to include stats that will ultimately dilute the significance of the stats, you don't include games like LSU versus Louisiana-Monroe, or Nebraska versus Western Kentucky.
We also looked at the numbers over a period of five years. Any statistician will tell you that in order for any study to have relevance, the data you're drawing on needs to be "robust," – there have to be enough numbers in order to make sure you're not just looking at temporary trends. Including the stats for every BCS game over the last five years helps us be a little more certain of our conclusions.
Below is a snapshot of the various categories we looked at, and the ultimate outcome in terms of correlation. In articles throughout the rest of this week, we'll break down the data and try to discover what it is telling us about stats that matter, and we'll relate those stats to UNC football in recent years.