I got an email from a professor today asking if I would help him with a statistical analysis project of his that is essentially an NFL Power Rankings system. My task would be to compare the results of the professor's pretty simple performance model - it uses only 4 inputs based on play-by-play results - to other similar statistical as well as anecdotal power ranking systems for NFL teams. You can check out the model's output each week in the Wall Street Journal in print in NYC or on the Internet World Wide Web here.
I think I am going to say yes and help out, but only if I am granted access to the model and fix the flaw that caused the GMEN to fall from 4th to 20th this week. Oh, wait. Nevermind. That wasn't the model's fault. That was an o-line that got torched, a d-line that got dominated, a non-existent linebacking core, a potential team cancer at running back (sound familiar, Tiki?) and an injured Kevin Boss.
Ranking highlights...
Rank Team Points
1 Colts 3.05
4 Texans 2.67
7 Jets 2.00
12 Patriots 1.41
13 Falcons 1.39
20 Giants -1.05
32 Panthers -3.68
Thoughts/rants on ranking highlights...
- How 'bout them Texans?
- That 7/12 spot has gotta sting half the readers/participants of this blog
- The 7/20 spot would hurt more if the GMEN hadn't dirtied Sanchez in the preseason
- Good to see the team the Giants beat in week 1 is dead last. I said after the opener that the Panthers were probably going to be the worst team in football this year.
A win last week would have put the Pats somewhere closer to the Dolphins, but alas. This is quite a ranking system where the 0-2 Boys are ahead of a lot of very good 1-1 teams.
ReplyDeleteVery interesting. Is this his first year doing it? I'd be cool to see some back-tested results to see how they align with the more traditional power rankings. It is weird to see the Seahawks at 8 and teams like the Ravens, who I think are very legit, at 24. Can we get an understanding for why offensive play success rate is the most heavily-weighted input? Sweet HW assignment C9
ReplyDeleteI would think that plugging different inputs from each team into a model after just two weeks would be slightly misleading, I'll be interested to see how this looks beyond week 8.
ReplyDelete