How Hard is It to Develop a Viable Algorithmic Trading System? (Part 4)

Do you realize this has been a whirlwind exploration into advertising, marketing, epistemology, and scientific method?!

We started with Davey’s claim about 100-200 trading ideas being required to generate a tradable system.

Then we thoroughly analyzed the idea that strategies in the public domain will not be profitable.

We talked about the impact of the “big boy” institutions with all their funding, personnel, and technology (i.e. computing power, lightning-fast networks, high-frequency trading platforms).

We talked about the impact of good-sounding ideas and the limits of our ability to understand the actual veracity of such claims.

In the final analysis, I think reasonable doubt abounds with regard to many things covered in these last four posts. Yes, strategies in the public domain should be low-hanging fruit for the institutions who can easily develop them.

But we don’t know if this actually happens.

We don’t even know how many institutions are fully staffed with financial engineers, quants, and an overload of computing power.

We certainly don’t know how much capital these institutions control nor how much capital is sufficient to exhaust the profitability from a trading system.

So I say give it a whirl! If you have the wherewithal then start with something simple that you can find in a book or on-line. Optimize to understand the parameter space and get an idea if the strategy is capable of robust results or only fluke occurrence. Walk it forward to see how the out-of-sample results look. If it survives those steps then incubate it by paper trading to monitor it. Finally, start live trading in small size…

…all the while, continuing to develop other systems to compliment or replace when need be.

This is how business gets done.