Actually, I was in college doing a math degree and I didn't have any money for opening brokerage accounts/data feeds.
I was using forex demo accounts to practice trading in my spare time and needed to be able to test indicators that I was interested in on real data.
I ended up just generating 'market data' myself using really high level probability theory.
That other post describes the type of functions I used.
I was really interested in stuff like figuring out if ideas like
bollinger bands
shifted moving averages
cubic spline
regression functions
auto-correlation
machine learning type stuff
residual analysis
bounding functions
conditional probabilities
EWMA and variants
had any kind of real edge in terms making money in the markets.
It turns out that knowing about this stuff does give you an edge, but not in an obvious way.