Things My Algo Taught Me

This week I finally automated my scalping strategy. Yes it uses Bollinger bands. Yes it's a reversion to the mean idea. No I won't reveal the specifics, so please don't ask me. What I do want to talk about, however, is the idea of algo trading itself. For a purely discretionary trader like most of us, algorithmic trading is a brand new frontier and I wanted to share my initial thoughts on the subject. Those of you who have been trading automated strategies for a while, please write in offer your insights. I would love to hear from you.



Whether we like it or not algo trading is the future for both institutional and retail accounts. I don't think computers will ever replace human analysis completely, but just like in in every major field of life from aeronautics to medicine to agriculture to transportation they will become a crucial component of economic activity. So here is a list of things my algo taught me so far

1. Train Yourself

The first thing you will want to do with your algo is to override it. That's fine and totally natural. As discretionary screen based traders we simply don't realize how much "art" is involved in our day to day trading process. Even the most steely, disciplined traders will change their "ironclad" trading rules on the fly in reaction to whatever is happening on the screen.

So go ahead, screw around with the algo, take profits early, terminate trades arbitrarily, override your entry criteria. In short let emotion take over and get it out of your system. After a few days of second guessing yourself and losing money you will finally be ready to let go. Like an anxious parent watching his baby take a few steps on his own, you will finally start to trust the process.


2. Trade with Real Money

I know the standard advice is to trade your algo on a demo, but I disagree wholeheartedly. Trading a demo is like playing HALO 3 and thinking that you are ready for war. The moment you hear a shell explode near you in real life you'll pee your pants. Mistakes WILL happen on the algo (in the future I am going to devote a whole column to all the things that can go wrong) and you need to learn from those mistakes quickly. There is no better feedback loop than watching your account get sliced and diced by the computer. You will become battle hardened in no time.

3. Understand the Model

Ask yourself exactly what price behavior you are trying to model. Ask yourself why you think this behavior may have an edge in the market. Stop obsessing about the indicators. Indicators are simply tools that help you describe your psychological insight mathematically, so changing the parameters of your SMA from 20 to 25 will not help you at all in a rangebound market. Always take an eagle's eye view of your strategy and be merciless with your assumptions because I can assure you, if you are not - the market will show you no mercy.


4. Let Algo Be Algo

Once you are generally content with the logic of your strategy, let the thing trade. Just as we must learn as parents to bite our tongue and let our children discover for themselves the rules of life, so too with the algo we must let it trade for a while to discover the rules of price action. Besides, you need at least 200 data samples to even begin to draw any meaningful conclusions from your ideas so sit still and let it operate on its own. After all that's the goal isn't it?

source: BKForex Advisor