Welcome to the Power Rankings Consensus.
This is a weekly series which aggregates a bunch of subjective (opinion-based) and objective (numbers-based) power rankings from across the internet, then averages and analyzes them.
This gives a wider perspective on what's going on in the NFL. Also, I like numbers and graphs.
Enjoy
Quote
I'm only including one quote this week:
The 49ers aren’t a good football team, but Colin Kaepernick is earning himself a job in 2017 with his stellar play in the past month. It may not be in San Francisco, but he’ll be a starter next season with a quarterback-needy team giving him a shot.
Fox, who ranked the Niners 28th.
Subjective Polls
This section gathers rankings from a number of sources which base their ordering on non-data-driven sources. I present a few ways of looking at the data:
An overall average, which gives an idea of which teams are rated high and low, evening out individual biases. The colors on this graph are green for teams with an average rank in the top 8, yellow for the next 8, orange for the third 8, and red for the last 8. Note that there are not necessarily 8 teams in each quarter.
A chart sorted by Standard Deviation. STDEV is a way of measuring how much disagreement there is in a data set - the lower the number, the more agreement there is. If all sources ranked a team the same, its STDEV would be zero. A high STDEV means that the critics don’t agree on how good a team is.
A chart of poll STDEV. This chart measures how close each poll is to the average. Lower numbers are closer to the consensus, and higher numbers have more controversial rankings. A graph of poll STDEV over time. This is a pretty messy graph, sorry.
Link to Subjective Average graph
Link to Subjective Standard Deviation graph
Link to Subjective Poll STDEV graph
This chart measures how much the polls agree with each other (a smaller number is more agreement)
Link to Subjective Standard Deviation of Polls graph
Objective Models
This section gathers numbers from sources which base their rankings on data-driven algorithms. There is quite a lot of variation in which numbers each site prioritizes, so these rankings tend to have much more disagreement than the Subjective sources, especially early in the season.
Also, since there is no data before the first games are played (unless you count the preseason, which I don’t), the Objective models don’t kick in until week 2.
The charts are analogous to the ones used for the Subjective data.
Link to Objective Average graph
Patriots are on top here, not the Cowboys, despite more sources ranking the Cowboys #1. Team Rankings has them at 8, which drops the average a lot.
Link to Objective Standard Deviation graph
Link to Objective Poll STDEV graph
This chart measures how much the polls agree with each other (a smaller number is more agreement)
I tried plugging in the numbers from Beatgraphs' other two models (iterative and weighted) to see if that would bring them more in line with the consensus. They're actually farther away.
Link to Objective Standard Deviation of Polls graph
Difference in Averages
Link to Difference in Averages graph
This is the subjective average minus the objective average. A team at the top of this table, with a positive score, can be considered underrated by the subjective polls, or overrated by the objective models.
Conversely, a team at the bottom, with a negative score, can be considered overrated by the subjective polls, or underrated by the objective models.
Acknowledgement
Thanks to John Crean at Bolts From the Blue, from whom I got the idea to do this. I’ve taken it a slightly different direction, but only slightly.
Sources
Subjective
Objective