What defines a robust system?

I can’t stress this enough; You need to have a robust system.

What is a robust system? A robust system is a trading system that has been validated, that works reasonably well over various markets and that the end user, the trader, can implement on an ongoing basis.

The first question I ask people is, why does your strategy make money? If you don’t understand why your strategy makes money, you’re going to find it very difficult to stay with your strategy when it goes a little bit pear shaped –which every strategy will at times. If we go back to the easy version of the momentum or trend following strategy, we make money because we cut our losses and let the winners run. You create a positive expectancy because you make a lot more money on the trades that win, than you lose when you have a losing trade.

Mean Reversion

With Mean reversion, I use the analogy of the rubber band. If prices stretch a certain distance and then stretch a little further they’re going to tend to snap back.

If I was to ask someone a question, “Why does your strategy make money?”, and they said “When the three day moving average goes across the ten, then the moving average in the Bollinger band is here, and the RSI is here, and volume is big, and the moon is up”, that’s not a reason why your strategy is making money.

Robust Strategy

A robust strategy is a strategy that works reasonably well most of the time –remember strategies will go in and out of sync with the market, that doesn’t mean it’s broken. It will go in and out of sync with the market. A robust strategy is one that will work across a large variety of stocks or instruments. For example, I trade the exact same strategy in both the top five hundred U.S stocks, the S&P 500 and also in the top five hundred Australian stocks. I use the same rules for every stock in both those markets. Every parameter setting is the same. Now that’s a robust system. There are a thousand different stocks there and we use historical constituent data. Meaning we completely remove survivorship bias and we can go back twenty years and we can see that strategy works perfectly well. There’s no optimization on any individual stock, there’s no optimisation on any particular market, and it’s the same rule. That tells me the strategy is very robust.

Non-Robust Strategy

To give you an idea of a non-robust strategy –and this is the trap that a lot of people fall into. And again, they tend to fall into this trap because they want something to work exceptionally well all the time, so they will tend to optimize or curve fit. Good example is a Gold ETF strategy that’s on the internet. Basically it buys the Gold ETF if it rallies more than 1.5% on a day –it will buy on the close. The next morning it will exit the long and sell short, and then get out of that position at the end of the day. Pretty simple rules and it works exceptionally well on this ETF. However, if you then take that rule, if that was a robust rule, we should be able to apply it to the S&P 500 constituent list, and it should give some kind of positive outcome –but it doesn’t. It’s a terrible system. So that’s not a robust strategy. It might be a pattern consistent with that particular stock but I would never trade it because it’s not robust. If it’s not robust it will tend to fail in the future. The more likely success over the longer term is to find a robust system.

A robust strategy must operate over a very large sample of trades. So for example, one of the main reversion systems that I trade across Australia and the U.S. I can trade it on the Russell 2000 and it still shows the same profitability. It does about six hundred, seven hundred trades a year. Going back fifteen, twenty years, you get a lot of trades and you still get the same kind of equity growth that you have with the robust system. As opposed to a strategy that’s curve fitted. You might only get twenty trades in twenty years, that’s not robust.