One of the best trading resources that I recently discovered is Andrew Swanscott’s podcast called Better System Trader. Even if you are not interested in systematic research and just want to trade discretionarily, the trading insights from the interviews are worth a listen.
One episode I found very valuable is an interview with Art Collins who is long time systematic trader in US stock and bond futures. Art wrote a book, called Beating the Financial Futures Markets which I have yet to read, but his analysis of what makes for a viable trading system really impressed me so I thought this week I would share his ideas with you.
Before all else, Art makes a point that I’ve heard over and over again from many different traders. The single most important aspect of the system is that it be in sync with your personality. If you are like me and like constant action then trading 100 times per day on a 1-minute chart is perfectly fine, as long as you adjust the system to the reality you’ve chosen. If you are like Kathy and think that such an approach is utterly ridiculous and prefer to make 2-3 well-chosen trades per day using the four-hour chart – that fine too. (I would rather get a root canal without anesthesia, but to each is own.)
That being said, Art has four key metrics to judge a system.
Does it make sense? Do you understand the underlying drivers? If you do not understand what the system is doing you will abandon it at the first sign of trouble. Generally, as I’ve noted many times before there are only two types of trading systems – continuity and mean reversion. Systems will naturally underperform in adverse market regimes, but If you have a favorable market environment (trending) and your continuity system is not performing you need to quickly assess what’s wrong and to do that you need to know how the trades work.
Don’t Optimize. Don’t Tweak. Don’t try to avoid the pain. Accept the drawdown because if you don’t it will only get worse. So if you are looking at a series of parameters make sure that if you chose a slightly different one the results will not be much different from all the other parameters. If they are that means your parameter is less than worthless because it only works on a particular set of data in the past.
At very minimum, the system must work on related markets. For Art that means that if the system is designed for S&P it must also work on Nasdaq and Russell. For us, in FX we need to make sure that the trade idea works on several related pairs, not just one. Earlier this week I had a system that looked very promising but when I analyzed the underlying data I realized that GBPUSD was responsible for 62% of the profit but it comprised just 16% of the trades. My new version was much better balanced with no pair accounting for more than 25% of the profit while comprising 16% of the portfolio. That’s the kind of distribution you want because that means you are capturing repeatable price behavior rather than one-off action.
And this is perhaps the most important and overlooked aspect of system analysis. Make sure that the bulk of your profits does not come from a very narrow time interval because then it’s a function of luck rather than skill. Since I day trade around the clock with fixed stops and losses, I avoid that problem by creating as much uniformity in my trades as possible. But if you trade on longer time frames with variable profits and losses you should study your results very carefully to make sure that they are not skewed by one or two lucky big trades.
Lastly, Art says that one of the best ways to analyze the robustness of a system is to divide the total profit by maximum drawdown – something I’ve intuitively done for years and prefer much more than the traditional Sharpe or Sortino ratio measures. But even here you need to be careful. If your system has massively large stops it could provide you with a very unrealistic picture of its robustness. For example one of the best traders in my room had a “return on account” (that’s what this ratio is called) of more than 10. She was up 22% on equity with a drawdown of only 2%. But that’s because the system was trading with massive negative skew (the risk-reward was 1:5) so the losses were rare and provided a false sense of security. Fortunately, she wasn’t fooled by the data and traded at very low leverage to prepare for any large losses that could come like an avalanche. Generally, the return on account of 2:1 or better is a sign that you are doing things well and a much better way to assess the risk of the strategy than the simple risk/reward ratio of any given trade.
I’ll be in Madrid next week at the annual Forex Day show, so no column next week, but come say hi if you are there, it would be great to meet everyone at the show.