BEFORE YOU TRADE YOUR MECHANICAL SYSTEM—A CHECKLIST

Posted by Art thoughts on August 13, 2016 in bids and offers

You believe you’ve done the necessary work in creating a mechanical trading system. Your results look good. You’re determined to follow your rules without question or deviation. Still, there’s a nagging feeling in the corner of your mind. Can I really trust my creation or did I miss something? And how would I know if I did?
We’ll never know for absolute 100 percent certain that our system will make money in real time as it has done in the theoretical past. For some of us, that can be a passing dark thought we accept as the cost of doing this particular business. For others, it can be somewhat paralyzing. Is there anything that can give us reasonable reassurance? The following checklist should help.
1. Did I test far enough back?
The short answer is you need to examine at least five years. Something that works in 10 total is even better and something that works in 20, better still. More validation is always better than less.
2. Did I account for varying market environments—bull, bear and stagnant?
3. When I optimized (compared market variables), did I settle on numbers that existed in a field of good numbers as opposed to a stand out diamond-in-the-rough?
To meet this requirement, go to your second best result if the best is an outlier in an otherwise losing neighborhood. If the second best doesn’t comply, go to your third best. And so on.
4. Does my system work in more markets than merely the one I’ve chosen to trade? Particularly, does it hold up to a reasonable degree in closely related markets—the S&P and the Russell, for example, or soybeans and corn?
5. Did I make sure that my overall profits were not coming near-exclusively from a narrow time frame such as a market crash environment? Are they reasonably evenly distributed throughout my data field?
6. Do I have enough capital to trade it?
Your worst drawdown is the clue—figure you’ll need at least one and a half times that amount.
7. Am I determined to see this through to the point where the long run probabilities have a chance to play out? Or will I lose my nerve after a string of losses—maybe far less than the worst string I saw in my data?
8. Am I ABSOLUTELY CERTAIN that I will do no improvisation or deviation from my rules?
A mechanical/discretionary hybrid will almost certainly give you the worst of both worlds.
If you answered yes to all but one question, go back and remedy the offending component. All of the above are essential for high probability of success. If you answered yes to all, chances are, your system is suitable for a real time launch.

One Comment

Comments are closed.