The rating system I have developed for soccer teams is a dynamic generalized linear model (see eg. Multivariate statistical modelling based on generalized linear models by Ludwig Fahrmeir and Gerhard Tutz) based on the Bradley-Terry model for paired comparisons, which is also known as a logistic model. The basic formula relating ratings to winning probabilities (ignoring ties) is
where P is the probability that the home team wins, 
is
a parameter representing the home field advantage (specifically, its the
log odds for a home team victory when the two teams are evenly matched),
and 
is a scale parameter chosen so that a rating difference of
100 points corresponds to a probability of 2/3 of victory for
the higher rated team at a neutral site, ie.
and 
are the
home team rating and the away team rating, respectively.
Here is a brief table of winning probabilities for the higher rated team, ignoring home field effects:
Diff Prob 0 0.500 100 0.667 200 0.800 300 0.889 400 0.941 500 0.970 600 0.985 700 0.992
For the 1995 NCAA men's and women's Division I teams, the home team wins about 60% of the time, which corresponds to
To put it a different way, the home field advantage corresponds to a rating difference of close to 60 points in this system.