hdot Image and File Hosting hdot - Upload File xerol.org Homepage xerol.org News/Blog Xerol's Music

hDot Image and File Hosting

Blog NFL Moon Rankings
Posted on 23:05:42 09/21/07 by Xerol

Share this!

Share

This is just something silly I'm doing for fun and math. Mostly for math. And somewhat for fun. Fun and math.


The formula basically takes into account two things: the ratio of points, and the rating of the opponent going into the game. So shutouts don't get wild point ratings (and divide by 0 errors) 50 is added to each score. So a 10-0 game gets 1.2 points (60/50) for the winner and 0.8333... for the loser. A fourth of the opponent's rating is added for a loss, and half of it for a win, so wins in general get about 0.25 points more. Throw on a modifier for the previous rating (which decays over time) and you get a rolling rating that I think should be a pretty good measure of performance and expectations. I took the previous season's record as the "score" for "week 0", so everyone started with a somewhat relevant rating.

So I threw the scores, opponents, and whatnot into a Google Docs spreadsheet. (Check out Google Docs; it's like a free online version of Office. Not fully featured by any means but great if you're on the road and don't have a spreadsheet program on you.) The rankings work surprisingly well, although the strength of schedule modifier (i.e. the opponent's previous rating) is weighted a little too heavily.

As you can see here the Bills, just by being scheduled against the Patriots, pick up about a dozen places before the teams even take the field. There's issues here, but it seems like it should balance out: messing with possible scores shows the Pats need to win by about 21 to maintain their rating, while the spread on the game is 17. The same tends to hold out for most other teams, with the required point spread being about 25% greater than the actual betting line, but generally in the same direction.

Okay, here's where it gets real nerdy. It still needs tweaking, but the total ratings for the week tend to total between 32 and 33, i.e. just over 1 point per team. Which makes 1.0 just about the median value, with 14 teams above it after 1 week and 17 after two. As the season moves on and more data gets put in, I'll be able to find a good line to mark off playoff contenders, but I think 1.1 is a good estimate.

The full spreadsheet can be viewed at http://spreadsheets.google.com/ccc?key=pYyk-Rw6YPFOuJu4dK9CTbw&hl=en.


Comments

Site Navigation
Xerol's Music

Code (c) 2006-9 Xerol

Contact: xerol@xerol.org - Put 'hdot' in subject line

Total hits since June 2006: 11777263