http://www.peterwebb.co.uk/probability.htm
is a pretty good layman's guide to probability theory.
He puts in something about forecasting correct scores, but one obvious difficulty with his system is that it looks only at how many goals the two teams in the example score - obviously it should take into account the number of goals they concede as well. But you should beware of averaging out the figures.
Last season in the EPL, home teams scored an average of 1.46 goals per game, while away teams scored an average of 1.02.
So lets say that team A averages 2 goals scored per game at home, and team B concedes an average of 2 goals per game. Does that mean that team A can expect to score 2 goals against team B? No; team B is conceding more than the average number of goals in its away games, and team A is scoring more than the average in its home games; if the teams haven't changed, team A could expect, on average, to score 2.76 games in a game against team B (this is (team A's goals/average for the division)*(team Bs goals/ average for the division)* average for the division)
Use this method, however, too many, and you'll go back too far to a time when the teams were different - Chesea and Arsenal are different from what they were a few months ago because of Gallas and Cole changing clubs, not to mention all their other transfers.
Use too few games and you run the risk of having a freak result - particularly one high scoring game - skew your figures.
Nonetheless, if you are happy with using goalscoring expectations to forecast your results, this is what you should do - if you have Microsoft Excel.
continued . . .