An Alternative Ranking for College Football
“The League Table Never Lies”
I am of a vintage where I remember Alan Hansen saying that phrase on Match of the Day. His point being that over the course of a 38-game season, your league points fairly reflect your position in the league. If you were in the relegation zone at the end of the season, you probably deserved to be there.
It is difficult to apply this principle to CFB, as the season is only 12 games long. A drop in form can destroy your season (ask Penn State). Also, everyone does not play everyone in a conference, before you even start to consider the FBS as a whole. Regardless, we still want to know who the best team is and we have the AP poll voters produce a rank. Towards the end of the season, the College Football Playoff Committee weigh in and select teams to enter the Playoffs.
Whilst some science is used in this process there is still a high level of subjectivity. To a spreadsheet junkie like me, this is highly unappealing! As such I set out to create my own version of a league table. Could I produce an alternative ranking that doesn’t lie?
Where to Start?
I first had to decide on what variables were important to me. I had to keep in mind the true homework question “which team is the best”. There is a plethora of statistics out there but I needed something quite easy to work with and easy to obtain on a Sunday morning.
I decided that to me, “the best” meant the team that was winning the most and winning well. I broke this down into the following variables:
Win percentage – self-explanatory, teams with more wins are doing well…
Average Points scored – a team that scores more points per game has a strong offense. The averaging is important to even out lopsided wins against weaker teams.
Average Points Conceded – as above, this gave me a sense of how strong a team’s defence was. Averaged again for consistency.
Margin of Victory – I wanted to then look at the difference between points scored and conceded to get a feel for how dominant a team was in their games.
Strength of Schedule – I wanted to add a metric to account for the relative strength of schedule. I settled on the BCFToys.com GLS metric ranking. This was easy to obtain and a good proxy for what I wanted. GLS is a Strength of Schedule calculation which asks “how many losses would a good team expect against a team’s schedule?”.
I normalised the GLS ranking and then created a formula and weighted the various factors to create a points total. I weighted the PF and PA by 10% each, the margin of victory by 20%, the win % by 40% and the SoS by 20%. I could have chosen any variation but settled on this for various reasons.
The Results – An Alternative Ranking
The outcome gave me the following top 25 for week 10.
Shortcomings
The above is a simple model and with simplicity comes shortcomings.
I am conscious that the formula relies on the SoS rankings which is an external calculation rather than something the team “does”. However, I thought it important to add this type of measure given everyone doesn’t play everyone. There needs to be a balancing factor.
I am aware that there are far better metrics for offense and defence out there (EPA for example) but again as a starting point I used PF and PA as a proxy as I needed something easy to use in a league table.
End Result
As a league table equivalent, I think this method works. But as I said at the start of this article, a league table across 12 games will not always truly reflect how good a team is. Luck and circumstance will play their part over such a short season. But if you win your games against good opposition and look dominant in doing so, you will find yourself at the top of the pile.
Picking Playoff Contenders
If the above is how I would think about rankings, how would I think about playoff contenders? I have said before I like the idea of ranking teams in some sort of league table format and then selecting a fixed number from each conference. Whether that is 4+4+2+2+3 or something else is open to debate. But for a spreadsheet guy like me, a league table never lies!
The CFB Playoff Committee will soon release its first cut of its playoff selections and we will see how the AP rankings compare to their selections.
My Top 25 vs Top 25 Polls (week 10)
Below is how my table compares to both the UK & Ireland Media Poll and the AP Poll.
There are 3 teams not in my top 25 that feature in the others. My rankings had Missouri at no 35, Memphis 28 and North Texas at 26.
