2011 NFL Power Rankings from Around the Web: Week 6

Subjective avg, Objective avg, Composite avg, STD DEV, red and green boxes…..what does all of this mean??

I really appreciate all the work that goes into these posts. Keep up the good work.

I love what you’ve got going here. It saves me from having to look up all of those rankings for my own curiosity, and shows me some I didn’t know existed.

Also, thank you for including BeatGraphs in your numbers this week. I noticed that you chose to use the Weighted method rankings. I would suggest using the Standard method numbers though, as one game can tend to skew the Weighted graphs. For example, the beatdown SF gave to TB. If TB turns out to be a good team and SF doesn’t, it’ll be nearly impossible for TB to wipe out that loss on the Weighted graph.

Best anywhere. Nice work, as always. If I were a betting man, I’d try my luck with the subj v. obj chart in Vegas.

When does DVOA eliminate their pre-season projection component (too lazy to look it up)? I’ve always thought that was the shakiest part of their early season rankings, and I wonder if it’s why it’s such an outlier compared to the other obj rankings.

Subj rankings looks like a big groupthink… not a whole late of variation between columns up and down the chart.

to my MBA stats teacher, he loved it. He always uses real data in his discussions and said this one might make it into a lecture. Keep up the good work.

YAY for cream’s smartness and green box percentage scale godness!!!

Through the week I usually go around and check the varying things myself but having the aggregation available is a major boon, so big props for your hard work every week.

If I could make a minor suggestion, I notice you added BeatGraphs (great ranking site btw) but there’s still a 3:1 ratio of subjective vs. objective rankings. This causes a certain heavy weight to the few number of objective rankings there are.

In the subjective rankings you’ve got, say, the 49ers who’re ranking between 7-9 with blips of 12-14 that don’t really skew the average much (but are represented in the standard deviation charts), whereas in the objective ranking a 10th place FO puts Minnesota up to a 19th ranked team when the concensus from all others is 22.6.

The subjective rankings are really good at giving a snapshot of what the consensus of a lot of places are thinking, whereas the objective rankings don’t give you as much of one. There are a whole lot of really good objective rankings available that could help give a larger perspective on what the “stats guys” around the internet say, like Advanced NFL Stats’s predictive power rankings, Massey Ratings, The Dunkel Index, and others.

I don’t wish to add any extra unwanted workload or anything, I know how much work goes into these sorts of things (particularly updating it weekly, good heavens). Just throwing those out there if they are of any use to you. .

I’m not sure the standard deviation chart is actually displaying teams that have the most variation in how they’re ranked from site-to-site.

Basically, teams in the middle of the rankings will always be subject to more variation (not just because they’re perceived as tougher to rank) because there are more slots available for them to deviate towards.

For example, the Packers cannot range from 1 “up” to 30, only from 1 down to 32. The odds of ranking systems differeing in 31 full units is significantly less than the odds of a team ranked 16/17 being ranked 1 or 32. In essence, the teams ranked closer towards the bounds of the rankings have less realistic - in the mathematical sense, not just the practical one — units to which their rankings can differ.

This means that a team ranked 3 in one poll being ranked 8 in another is a more mathematically robust difference than a team ranked 16 being ranked 21 in another. This should be reflected in the standard deviation chart. One way to do this would be taking the maximum deviation that any team exhibits in a week as a baseline. Say, for example, that some team differs by 8. Each team can then have their deviations scaled in a way that apapts to the number of units they can move to on either side, capped at 8 in any direction.