estimating football results using Poisson distribution

I currenlty doing my disertation at university and am creating a model that predicts football scores ive done alot of the coding itself but now im trying to account for games that may end in a draw, im unsure of the best way to do this, based on my model and data i have used i can correclty predicts 80+% of home wins but there is failure when doing away with a succes rate of 42% of correct results. Im a little bit stuck its been suggested that a supervised learning algorithm could also be used as a way of predicting scores ive looked into naive bayes but am not getting very far with manipulating the data any help would be appreciated

all matlab and excel spreadsheet files ive used are attatched

[Luke wilkinson]

for context the theory in which the file of expected goals comes from a paper by M H Maher estimating association football scores :)