Sunday, April 8, 2012

Power to the people

Okay, are you ready for some Week 5 power rankings? No? Too bad. Tie your shoes. Here we go. (*edit 4/10/12 - I discovered a small error in teams that had played 6 games calcs that changed the standings just a little, as well as caught an erroneous DC result...)

The Good

1. Kansas City (2.167)
2. Salt Lake City (4.500)
3. San Jose (4.833)
4. Seattle (7.000)

The (not so) Bad (cutoff is 7.333)

5. Colorado (8.000)
6. New York (8.167)
7. Vancouver (8.500)
8. Houston (9.333)
9. Columbus (9.677)
10. Chicago (10.000)
11. Dallas (11.500)
12. D.C. (13.167)

The Ugly (cutoff is 13.500)

13. Portland (13.500)
14. L.A. (13.677)
15. Chivas (13.833)
16. New England (14.000)
17. Philadelphia (15.667)
18. Montreal (17.167)
19. Toronto (19.000)

To explain how I come up with these, it might be best to just post the excel spreadsheet. I'll try to upload it as a Google doc tomorrow, but, basically, I look at the teams Last 5 game form for both results and goal differential, and do a sort of fading weight. The equation is basically this.

Results Form = Last game points earned + Game before that * 0.4 + Game before that * 0.3 + Game before that * 0.2 + Game before that * 0.1.

GD Form = Last game goal differential + Game before that * 0.4 + Game before that * 0.3 + Game before that * 0.2 + Game before that * 0.1.

(I should note that for teams that haven't played 5 games yet, I do change the summing so it adds to a total of 2.0 given their total number of games played. This won't matter after a few more weeks, once Houston comes back form their vacation.)

I then assign a Rank to each team each week, weighting Results 2:1 with GD. So, each week, a team's rank is

Rank = 0.67 * Result Form + 0.33 * GD Form

Which sounds well and good, but we still need to look at each team over the course of the season to really get a feel for how good they've been.

I feed all that data into a tracking table that looks at how each team has been ranked each week and spits out stats, then use their average ranking to assign a "power ranking". Each ranking is based on the teams average rank throughout the season.

So far I actually think it's worked out fairly well. It's less "gut feeling" than other power rankings as it's actually based on results, but it still doesn't take into account how good a team is on paper. Still, I feel like performance is the best metric, and as the season goes on it should get better at telling us who's hot and who's not (~cough~ L.A. ~cough~).

No comments:

Post a Comment