Sample size for a sports betting strategy

I picked up a book called Simple Soccer by some Brazilian guy (he might be called Leandro Vieira, but I'm not sure). In there, he outlines a very simple strategy for betting on football (or soccer, if you're in America). Even though I didn't like the strategy at first, because it involved betting on favorites (and *only* on favorites), he claims a 5% return rate, which didn't seem too preposterous. So, I decided to give it a go.

Even though my results are even better than his (8.7%), my sample size so far is only 17 games. I'm curious on how many bets before we have a big enough sample size which allow us to conclude if the strategy is viable or not (say, with 90% certainty). The author claims 30 bets would do it. I'm not convinced. Also, would different sports need different samples? Soccer also has three outcomes - drawn games account for about 1/3 of the results. Would this make a difference?

I'm sure one of the more mathematical people will correct me, but basically, the sample size of 30 assumes that the population is normally distributed. Approximately 90% of the population, if it's normally distributed, will fall within +/- 1.65 standard deviations of the mean. The standard deviations within the sports are likely different, so you'll need a minimum of 30 samples per sport. This also assumes constant variance.