Probabilities are obtained by using the differential in "power" ratings of the two sides in question with HA (home field advantage) also taken into account. The power ratings are based on the previous form of each particular team with several performance factors taken into account and derived from a mathematical model that optimises and reduces the error of these ratings. The power ratings change from week to week depending on present as well as past results to best predict the outcome of each match.

This week my model predicts New York Giants, Atlanta and Oakland as favourites to win(>50%) with the first two of these sides playing away from home the market(Betfair) has them all as underdogs(>2.0). My model rates them as 64%,67%,57% chances of winning respectively.

One thing the model does not take into account is each team's line up from week to week. An obviously important factor not used in the calculations.