For the last few seasons, I’ve published my own version of the popular NFL Power Rankings that appear on a number of sites. On most sites, these seem to be calculated arbitrarily, whereas I’ve tried to produce something that adds at least a slightly more scientific approach. What follows is an explanation of the method, but if you just want to see the rankings, they’re at the bottom of the piece.
It’s worth noting, before reading further, that I’ve adjusted my rankings formula again, so if you’ve read my rankings in previous years, they’re not directly comparable.
Finding common games
This approach is made a little tricky by the way the NFL organises schedules, but not impossible. No team plays the same sixteen opponents across the regular season, but in each division, every team plays a similar schedule across fourteen games:
- Home and away against each team in their own division;
- Once against each team in a designated NFC division;
- Once against each team in a designated AFC division.
There are some differences in that schedule, as the combination of games played at home and on the road is different, and the final two games are unrelated, based against each team’s relative performance in the previous season, but it’s a good start. Thus, I base my Power Rankings on those fourteen common games, ignoring the two outside that.
As games are only common within divisions, I need a way to compare divisions. I do this by comparing the number of total wins in each division, as each division. Whichever division has the most wins is the strongest division (for now, at least). It’s the mark by which all the other divisions are marked. All other divisions are then assessed based on their relative percentage of wins. If the division with the highest number of wins has forty wins, and therefore has a relative strength of 100%, a division with thirty wins would have a relative strength of 75%.
Whilst a good start, this is still very crude. What if a division had fewer wins than another because the divisions it was matched up against were actually very tough? What if a division Was matched against two weaker ones, and so had an easier ride?
Well, this is relatively easy to solve, as each division is matched against two others. So, by combining each division’s relative strength with that of the two it was matched against, I can calculate an overall division strength, and I do this by adding the three values together. Thus, carrying the previous example, if the top division has forty wins, and it’s up against two divisions with thirty wins each, the formula would be:
1 + 0.75 + 0.75 = 2.5
The resulting figure of 2.5 is the division’s overall schedule strength. A higher number means a tougher schedule. This value is then, for each team multiplied by their record as a decimal value, where 16-0 would equate to 1, or 12-4 to 0.75. The resulting figure is the team’s final ranking.
The final formula
So, the final formula (when fully expanded) looks like this:
(D + A + N) * R
- D – Division strength
- A – AFC opponent strength
- N – NFC opponent strength
- R – Record
Whilst form is important enough to be in, I reduce its weighting because I consider the overall division strength and record to be a more reliable indicator.
So, without further ado, here are the 2015-16 NFL Power Rankings (as of week 16):
|New England Patriots||2.382||0.800||1.906|
|Green Bay Packers||2.853||0.667||1.903|
|Kansas City Chiefs||2.706||0.667||1.805|
|New York Jets||2.382||0.667||1.589|
|St Louis Rams||2.735||0.467||1.277|
|New York Giants||2.706||0.400||1.082|
|New Orleans Saints||2.382||0.400||0.953|
|Tampa Bay Buccaneers||2.382||0.400||0.953|
|San Fransisco 49ers||2.735||0.267||0.73|
|San Diego Chargers||2.706||0.267||0.722|
I’ll likely write a separate post about the various methodologies I’ve used to calculate the Power Rankings, and why I’ve changed them along the way. I’m also open to refining the formula and also to know if any of the teams look out of place in these results!
Wondering why I’m publishing this before the final regular season games? I’m doing it because I want to see how the rankings stack up against the week 17 games. Not entirely scientific, I know, but interesting all the same.