My Perspective on Algorithmic Trading (Part 4)

I think the biggest disadvantage to algorithmic trading is how difficult it truly is to find viable trading strategies.

This is the thesis I’m working toward, but I won’t get there today. I see much to unpack in the one sentence above.

I believe discussion of “viable trading strategy” can be multifaceted and lengthy with no definitive answer to follow. I’ve done a lot of writing on this topic. I think the blog mini-series starting here is still quite relevant. Another relevant mini-series begins here. At some point, I may review some previous discussion in light of new information I have learned. For now, the bottom line (mentioned in this second paragraph) is that any process used for development must give me the confidence required to stick with an unbroken strategy in the face of “normal” drawdowns.

While having a process to find profitable strategies is one thing, implementing the process successfully is quite another. At first glance, I can find plenty of links that suggest such success is readily available.

Many vendors sell black-box trading systems. These are undisclosed algorithmic strategies that I can purchase (or lease) for use. I won’t find out how they work, but I will be privy to their returns if I allocate capital to them. Some of these have the most compelling sales pitches around.

Black boxes are algorithmic trading systems, but I’m not a fan as mentioned near the end of this post. If a pitch does include backtesting data, then I would want to double-check since backtesting is vulnerable to numerous potential pitfalls (examples reviewed here). Just because performance numbers are printed does not mean they are accurate (and we can almost be certain they aren’t this). If a pitch does not include backtesting data then I would certainly want to test it myself to see how it performs. As black-box systems, I can’t see the code and I therefore can’t backtest them. This is a deal-breaker for me.

Technical analysis (TA) lends itself well to algorithmic trading when it provides unequivocal, programmable rules. I can do an internet search on TA books and find a plethora that have been written on the subject (one of which has made a 9th edition). I can run an internet search on “learn TA” and find information to keep busy for weeks in the form of educational courses, blog posts, podcasts, etc. Many educational services out there advertise teaching trading with TA.

My general criticism of TA breaks down into three main categories. I will discuss these next time.