Most math around economics is basically guess work. There's an assumption of normality of log returns that is mostly false in the tails. The tails of the return distribution tend to be significantly different than the tails of a normal distribution - however majority of "data driven" approaches are underpinned by this assumption. As a result most applications of hard math to predicting stocks have failed. This is also, funny enough, why put call parity falls apart in practice and often times there are significant opportunities for arbitrage in stocks and futures. The major greats in financial math, for example Ed Thrope, took advantage of a flaw in the pricing of contracts - and not the predicting of stock prices themselves.
The application of machine learning to finance is more or less useless in the retail space. I have book called Advances in Financial Machine Learning which is the seminal work on this. Not only does the author go into mathematically complex ideas that are only really implementable at hedge fund scale, he also makes it clear there's virtually no room for this work in the retail space. There just isnt enough good quality data for a retail to do any meaningful work. You need millions of ticks of cleaned and processed data to have any hope of getting something like a neural net to be useful. Even then, you need to decide how you bin your ticks, what works and what doesn't, etc. It's a daunting task that needs a team of quants and data monkeys. You are one guy.
For me, charts aren't useful for MACDs and such. There are a handful of indicators I consider useful - namely indicators that tell me supply and demand characteristics of the market. I trade based on these, news, and probabilities. Since I trade options I can more-or-less define my risk characteristic for each trade. If the odds are good and the price is right, I'm in. Period. I just use charts to confirm there's something of substance (a major pivot off of a point the market is telling me is a demand limit, for example). Recently I've been profitable just watching Trump's twitter, so it goes to show you that you really need to be creative.
