This kind of thing is fine for solo-developers but it does not scale well to a team. When there's a testing framework in place, people's tests fit a standard pattern. I can go and look at someone elses test cases and I understand what they're doing to test it, because there's a standard way of doing it. There are methods in pretty much every testing framework for doing things like floating point comparisons correctly; with your way, you have to build up methods to do that yourself (i.e. creating your own framework anyway), or do that inline every single time.
