You probably know this, but an algorithm which is right 50% of the time is good if your algorithm is finding more value than it loses. i.e you might be right on 50% of 3/10 chances (paying 3.3-1) and wrong on 50% of 8/10 chances (losing 1.2-1). You win bigger on the wins, and lose less on the losses etc.
Not saying your algorithm does that mind, you'd need to stake it over 100 games and see if it is profitable. You would need to be better than 5-10% right to make up for the bookies margin mind, which is the big problem.