I'd say play around and see if you can predict the price - treat it like any machine learning/optimization problem - if you want to learn and have fun. If your goal is to make money, unfortunately it'll take a lot more. I've worked at a high frequency trading company, and now work at an algorithmic hedge fund, and you really need to scale horizontally to make money overall.

There's a lot I can talk about but here are some points:

- Most of the time you don't trade instruments in isolation. They are all correlated (even if inversely) one way or another , so movements in one will affect others. Having a myopic view of just one instrument will have too much unexplainable randomness.

- The other way to scale is having more data inform your pricing - i.e. don't just look at the prices of one or more instruments. Look at twitter, look at weather reports, news etc. It kind of suggested that in the article. The meaningful movements in the market are most of the time due to truths outside of the price itself.

- You have to decide whether you want to trade directionally and long term (like a hedge fund) and take on positions over time, or trade in and out of positions quickly (like high frequency trading firms) to minimize risk. You can be anywhere along that spectrum. The faster you are, the more myopic you can be and just react to current trading activity of other participants. The more long-term you are, the more you have to look at the big picture, more data, more instruments, making sure you have the right balance in your portfolio (whatever that means to you).

I'd be happy to expand more :)