As a hobby, I keep track of all Division 1 football scores
and use them to create my own personal ranking system.
Using least squares techniques, I adjust these rankings
to give predicted score differentials in each week's games.
Of course, no such ranking will ever correctly estimate all
game outcomes, but the least squares method gives the
ranking that will minimize the overall (total) error.
I typically have about 75% accuracy in picking games.

In the past I tried using an extra factor for games with lines,
or point spreads. This did not turn out very well, and it wasn't
until I abandoned it and went back to my previous system
that I got accuracy against the spread back over 50%.
In 2008 I am returning to my earlier ATS prediction system,
hoping to get back to my previous level of 60% accuracy.

These predictions are for recreational purposes only.
I do not spend money betting on games, and this site is not
intended to encourage people to gamble on sports events.
Rather, it serves as an ongoing exercise in seeing how a
simple mathematical model performs against the so-called
"experts" who charge outrageous prices for football picks.
If you do choose to participate in recreational gambling,
please do so responsibly.