The best “edge” available to traders is to trade with the trend: call it trend, momentum, whatever you like. The exact entry strategy you use probably won’t make too much difference to the results. I’ve written many times about how the best way to apply this kind of trend-following strategy to Forex is to stick to the three major currency pairs and use a “dual momentum” strategy of applying look-back periods of both 3 months and 6 months. I can use historical data that shows this method has worked well overall since the turn of the century. Yet when I ask myself how I came up with those periods – whether they are not just “curve fitted”, I reach an uncomfortable conclusion and look for a way around it. Stick with me, because if you trade in this way, these are questions worth asking and trying to answer!
In the early 2000s, trend-trading the four major Forex currency pairs (EUR/USD, GBP/USD, USD/CHF and USD/JPY) using a look back period of 12 months (which is traditional in stock investing) would have worked better than 6 months. By 2008, using shorter look-backs such as 3 months or 6 months started to give a better overall performance. The chart below shows how a simple long/short single look back strategy on these pairs would have worked from 2002 to date:
The above chart does not include trading costs or fees for holding positions overnight, it is just indicative of how four look-back periods would have worked. Another important thing which the chart does not show is the relatively inferior performance of USD/CHF on all look back periods since the SNB’s intervention in 2011.
Are there any answers to the dilemma of selecting the look-back period? There are a few possibilities:
Using the best performing look-back period. The problem is that there will be a lag. My own test shows this produces a worse result than averaging all four periods if a 3-year “walk forward” is used to determine the look-back period. How to know what look-back to use to measure the look-back? It is ironic, because you end up with the same problem you were trying to solve!
Use two periods, e.g. the price must be the same side of both a 3-month and 6-month look-back. There is some evidence this works better, but it does not really solve the problem outlined in 1. as you then must decide which time periods to use for the dual momentum look backs.
Use relative volatility to determine the length of the look-back period, where in a period of higher volatility, you use a longer-term look-back. This is an interesting potential solution which I have only just begun to test. The logic is, if there is higher volatility, the price is more likely to give a “false” momentum-based reading, which sounds persuasive.
I will have more to say about a volatility-based dynamic look-back soon.