July 1, 2014

You want to automate your trading strategies. Why not! You have out there, more than enough available software platforms, enough brokers, enough computer power, enough stocks to chose from and you have studied the markets for what appears like ages. After all, it's just an extension of what you have in mind and stuff you probably have seen or have been doing for years at one time or other.

You simply code your discretionary trading methods. Doesn't it sound easy? Especially for someone with an extensive comprehension of markets and trading techniques. At its very basic, is it not the application of common sense to solving a problem? General framework: if stuff x true then buy (cover); if stuff y true then sell (short); if between stuff x and y hold. There you are, a complete trading script with its 3 main trading decisions. It's not a question of having an opened mind. If it's not opened yet, you're a desperate case. It's a question of what is your trading methodology. Can you formulate what you have in mind in code?

Your main interest in stock prices is the part that makes you a profit. A mathematical expression for this is ΔP > 0. That's what you're looking for and running after for every single trade: that little expression. How complicated can it get?

Millions, and I do say literally millions before you have tried to design profitable trading systems. They all started the same way. They all had to redo what others had tried and tested before them, just like you probably did if you read this far. A lot of systems you've tested, or designed on your own, fell below your expectations and were shown to be not that profitable (a euphemism for poor performance).

Strategy designers tend to often deceive themselves by not designing viable systems; by not completely doing their homework. It's easy to do. One simply designs them from hindsight which more often than not will produce profitable systems. However, there is a major caveat: they worked fine over past data, probably on the single stock they were tested on, but are awful or break down going forward.

So you go back to the drawing board, again and again. It's a slow process, you try to build on the shoulders of those that preceded you, trying to re-invent what has probably already been done by someone else. It's only that almost none of those people put their research or code on the line, in public view. Designing a trading strategy is a continuous iterative process of: if I coded this or that, what would it do, what would be the result? When maybe at times a better approach might be: look, he or she's done that, and I can do better.

In designing trading systems you soon find problems. What you thought would be simple turns out to be not so simple after all. Concepts you relied on for years did not produce much when tested over years of data at the portfolio level. As a matter of fact, the more you coded stuff the more it produced about the same: underperforming the indexes over the long term. Not even being able to outperform the Buy & Hold. This is a harsh statement: you could have done absolutely nothing all those years and done better than your own programs (with all the implied know how).

I hear you saying: ah, you don't understand the way "I" trade using "my" discretionary trading methods. Sure, as long as you do not put them on the table, put them to the test, or show the results of some kind of simulation over say the last 10 years or so, how can I even guess the value of your trading system other than taking your word for it? I can backtest and put to code Mr. Buffett's trading methodology over its 50-year history. Analyze his discretionary trading methods, and show that he was not only right in doing what he did but also the why and how he did it with so much success. I can even write down the equations governing his investment methodology.


Created... July 1, 2014,    © Guy R. Fleury. All rights reserved.