Predicting Match Outcomes

I have a presentation to on predictive modelling and I was thinking of doing it on a game of premier league football (between any two teams). I need to find enough variables to make a large enough model so I thought I would ask the ja community for their opinions on what should be included.

For example, one would be the strength of the team or manager rating, whether the game is home or away, recent form, etc. Obviously the numbers don't matter at this stage, I will choose them later. Think of it like putting on a bet, what factors influence you to bet on one team or the other.

Also, admin1, what variables do you have in your betting algorithm?