The weights in the multiple regression analysis used to predict a team's "offensive score" are explained in that link after the 19 weekly Power Rankings: http://cfl.ca/page/cflca-mathematical-power-rankingsnotahomer wrote:Anyway I'd be curious as to how they build the formula that spits out their CFL rankings.....
Then they explain (see the quoted material below) how they use those to arrive at the Power Rankings:
_______________________________________________________________________________________________________________________________________________________________________
"Each week a Club’s 'weekly score' will be determined by calculating a Club’s 'offensive score,' using the formula above, then subtracting its opponent’s score calculated using the exact same formula.
Once a weekly score has been determined, the Power Rankings will be determined:
1. For the first half of the season by taking each week’s score and weighting them evenly (for example, the Power Rankings in week 4 would weight each week 25%, week 5 would weight each week 20%, and so on), except for the first week, which is based 50% on the regression score and 50% on last year's standings; and
2. For the second half of the season, after Labour Day, by weighting the current week 20%, the immediately prior week 10%, and all other weeks 70%."
_______________________________________________________________________________________________________________________________________________________________________
As for turnovers, it appears that fumbles and turnovers on downs are in the equation, but not interceptions. This might because the latter factor into the QB rating, which is given weight.
As for the rankings themselves, they look about right to me.