There is only one criteria, game margin. The computer varies the strength of the teams such that the sum (of the squares of (absolute value of (predicted games margins - actual game margins))) is minimized.
This is a standard multiple regression problem. Independent variable is the game margin. Dependent variables (outputs of the regression) are the individual team strengths and 1 home field advantage variable.
Mathematically, each game played is represented by an equation, involving 3 unknowns (2 team variables and 1 home field advantage variable) and 1 known (game margin).
Currently there have been 559 games played = 559 equations. 361 Iowa teams have been involved + 15 non-IA teams + 1 home field advantage variable = 377 total unknown variables. Since the number of equations is greater than the number of unknown variables, a statistical process known as regression is used to obtain the best available estimates for the unknown team strengths and home field advantage.
Once the individual team strengths have been found by the regression, predicted game margin = (home team strength + home field advantage) - visitor team strength.