What to do when your Trading System is not the problem
You think you have great entries and exits, and you have position sizes that you’re comfortable with. Yet when you systematize them, the hypothetical results are not critical to the extent that you can use them to manage risk — or raise any capital to trade.
So you go back into the dashboard of your simulator and you start the tweak the living daylights out of your entries. That doesn’t work. You tweak your exits. That does’t work either. You test smaller positions. That doesn’t work.
You begin to add indicators to your rules. They bring small but positive changes, but the overall hypothetical results still stink. You layer another 47 indicators on top of one another and you finally have something that will make money. But now the problem is that the system makes only 3 trades over the last 10 years. It basically needs to be Leap Year for you to put a trade on.
You’re about to scrap the whole file and start over. You’re feeling enormous buyer’s remorse because although the simulator isn’t necessarily expensive, it feels so when you can’t get it to work. You may become so frustrated that you want to quit.
The problem could be with your data. Yes, it’s important to have clean data in that you don’t want any bad prints to trigger trades in your simulator that will otherwise destroy your ethos when it would otherwise yield results that would look promising on the first pass. I’m talking about the universe of names (tickers) in the data that you’re pumping through the simulator.
You may need to rake the data before you run it through the simulator. Gold does not trade like Sugar does not trade like Apple. Why would you have them in the same data universe? You wouldn’t and that’s an easy one to understand. Suppose your trading rules worked better in one asset class that another? Or suppose they worked well for one capitalization but not another? Maybe you should consider segregating the small caps from the large caps…
Here you are double-clicking your mouse for 10 days trying to get your system to work and the problem is not with the system, but in the data.
You can rake your data before you run it through your system and create your own universe of securities to test. Here are a few examples:
–NYSE listed securities only
–Average Daily Trading Volume (ADTV) > 10MM shares
–prices between $20 and $75
Test everything at the portfolio level to make sure that your rules are robust, and try to keep your rules simple so that you can explain them very easily to a non-pro.