I also run a program to come up with predictions in the AFL.
But rather than using Excel,I use Basic.Only because it's what I learnt in my previous life.
What happens is,the computer crunches the numbers,and averages and weights the figures to come up with predicted scores for each team.
The inputted data consists of results from the previous 28 rounds.Why 28?.Mainly because it contains the previous 12 months of results.Had played around with the last 100 rounds,but found that old stats were contaminating the results.28 seems to be a good balance.
I just input the teams scores and who they played.
I also tinkered with using results to come up with a perceived home ground advantage,but found that this advantage was not a constant number and varied depending on the opposition.I now use a seat of the pants bonus system.Not very scientific but seems to work okay.

This year I am rating the players based on the stats they accummulated.This will hopefully give slightly better results than last year because it will allow for who's in and who's out.Early tinkering suggests only a difference of about 6 points for even the best player missing or returning,but as we all know,you still get nought when you go down by a short half head.

Other types of human intervention are needed to get accurate predictions.
Apart from the travelling and home ground issues,I allow for other things which reflect my experience in the school of hard knocks.
These include penalties for these burdens.

Grant Thomas -12.Kevin Sheedy -9.Dean Laidley -6.
Wanting to win it more +3 to +15.
Preferring to play rather than be best man at your brothers wedding +4

Many subjective assessments will give you a prediction that does not coincide with what bookies offer,and therein will lie your edge.And if you ain't got an edge,you ain't got nuthin'.