Earlier this month, I published the first part of this series, which explained how trading a “time series” momentum strategy restricted to USD and EUR currency pairs and crosses has historically performed far better that implementing a “best of” momentum strategy across a universe of currencies, at least over recent years.
In this second part, I will get into the details of the results of the back test I conducted, and show the differences in performance between the EUR pairs and the USD pairs, as well as examining how overall performance can differ with the application of certain filters.
USD vs EUR Currency Pairs & Crosses
I wrote in Part 1 that “Over a period of 6 years – from April 2009 to April 2015 – if you looked at the 28 most important currency pairs and went long or short of each every week depending upon its look-back periods of 13 or 26 weeks, the only currencies producing positive results were the EUR and the USD. Both currencies would have produced a return of 110% each based upon the 26 week look-back period (corresponding to 6 months). Using the look-back period of 13 weeks (corresponding to 3 months) produced a positive result of 161% for the USD and 82% for the EUR.”
Let’s take a closer look at these results, by conducting a 13 year back test on USD and EUR currency pairs and crosses (concluding in 2015), and working though the numbers.
4 Week Period
The first look back period was 4 weeks (corresponding to 1 month). The results are shown below:
The USD pairs achieved profitability, but would have been in a draw down for more than three years to date. The EUR pairs and crosses were quite consistently unprofitable. Overall, the combined result was very slightly positive over the period, by 19.46%. This period is really too short to use, although it seems capable of providing some profit over the long term.
13 Week Period
The results for 13 weeks (corresponding to 3 months) look much better, although it has to be noted that both sets were under a draw down for more than one third of the period. The USD pairs performed better, at 328.44%. The combined return for both was 485.58%. This looks like a good time period to use as a look-back.
26 Week Period
The results for 26 weeks (corresponding to 6 months) also look good, and possibly even better than the 13 week results. The USD result in particular looks excellent, with a fairly consistently rising equity curve, and a shallower draw-down during the earlier part of the back test compared to the 13 week results. The USD pairs performed better, at 317.43%. The combined return for both was 388.19%. This also looks like a good time period to use as a look-back.
52 Week Period
The results for 52 weeks (corresponding to 1 year) do not appear to look very good, although the results are considerably better than the 4 week results. The EUR result in particular looks interesting, with return of 196.83%. The USD pairs performed much worse, at only 49.89%. The combined return for both was 246.72%.
Analysis of Results
Time Period USD Performance EUR Performance TOTAL Performance Maximum Drawdown Longest Drawdown
One further consideration has to be accounted for: the transaction cost. There would be approximately 3,000 trades taken, which might account for a deducted of about 30% from the overall profit, and would increase somewhat the size of the maximum drawdowns and lengthen the longest drawdown.
Taking this into account, we can draw a few conclusions:
1. The results for the USD pairs look better than the results for the EUR pairs and crosses.
2. The form of the USD results look more logical, with each time period showing a profit, which seems to form a bell curve, peaking around 13 to 26 weeks.
3. It appears likely that this strategy can suffer an approximately 4 year drawdown, even with an excellent long-term performance.
4. The obvious ways the back test might be improved would be by sticking to USD pairs only, and perhaps taking only the trades that qualify under both the 13 and 26 week time period look backs, as a type of “multiple time frame trading” filter. The results for this variation are shown below.
The USD performance was 310.90%, comparable to the USD results for both the 13 and 26 weeks as standalone periods.
However the maximum drawdown was considerably lower at only -54.51%.
The longest drawdown was 237 weeks, which is quite comparable to the earlier results.
The methodology also lowered the total number of trades, thus producing a greater return per trade and lessening transaction costs.
Adding the EUR pairs and crosses does not improve the drawdown results.