Automated forex trading software makes it easy to test the historical effectiveness of different trading ideas, but how do we know when we have found a winning system? This is far from a straightforward task, and backtests can often prove misleading. In this article we will examine whether we can build a promising trading system using multi-day rallies and declines as trend continuation patterns.
Consecutive Day Rallies/Declines – Can We Trade With These Patterns?
Through strong trends, we will often see a currency pair rally or decline for a number of days in a row. The charts below plot the instances in which the EURUSD has rallied and fallen for N consecutive days in the past 10 years of trading. As N grows, the instances in which the pair has rallied or fallen that far shrinks quite quickly. There are two quite-different ways to use such information: we could choose to fade the initial move or chase continued trends.
We can test both ideas rather easily using automated trading software. In fact, we only really need to run one set of trials because our equity curves can be flipped to achieve the opposite result. Thus we will code some fairly straightforward logic into Strategy Trader and examine the backtest results.
Consecutive Days Up and Down Trading Continuation Strategy
Using Strategy Trader, we can import a readily available Consecutive Days up and Down Trading Strategy with the following trading logic and check the performance results.
Forex Consecutive Up and Down Days Continuation Strategy
Entry Rule:If the currency pair has rallied for N consecutive days, place a market entry order to buy on the next bar’s open. If the currency pair has fallen for N consecutive days, place a market entry order to sell on the next bar’s open.
Exit Rule:Trades are reversed with the opposite signal. Thus a long position will turn into a short position if there are N consecutive down days and vice versa.
Stop Loss and Profit Target:For purposes of these tests, we do not use stop losses or profit targets. Suffice it to say, however, any live version of such a strategy would need to use some sort of maximum risk parameter.
Backtesting our Consecutive Up and Down Days Continuation Strategy
We ran this strategy on the EURUSD, GBPUSD, USDJPY, USDCHF, USDCAD, AUDUSD, and NZDUSD pairs using their respective volatility values. We assume transaction costs of 3 pips on the EURUSD, USDJPY, and USDCHF and 4 pips on the others. Below are the hypothetical equity curves of said strategy run on 8 different values for our N consecutive day input.
Generated using FXCM Strategy Trader
Hypothetical trading results must be taken with a great degree of skepticism, as backtested results can be and most often are different than what would have been achieved with live trading. That being said, our paper trading results are rather interesting.
The chart and table above show that an input of 6 days would have achieved fairly substantial profits and reasonably good risk-adjusted returns. In fact the hypothetical drawdown was half of that of the maximum strategy equity—giving us a theoretical “Return on Account” of nearly 100 percent. Of course, we know better than to just accept results at face value. And the motivation behind this article is to examine exactly whether we can expect such paper trading results could produce a winning walk-forward strategy.
Results too Good to be True?
To analyze performance results, we need to understand why a strategy has or has not done well on our tests. In this case we backtested our strategies across a total of seven currency pairs. The first logical place to start is a breakdown of performance across specific instruments.
According to our currency pair breakdown, this strategy theoretically turns a profit on an impressive 6 of 7 currency pairs. Yet when optimizing across specific parameters, it is important to test not only for the best but also the stability of the best result. In this particular instance we’re interested in knowing how the strategy would have performed if we varied our input only very slightly—by 1 day.
The chart above suddenly shows that our performance falls apart when we change our input from 6 days to 5. In fact, 6 of 7 currency pairs now turn in at best a negligible profit and at worst a sizeable loss. After seeing such strongly positive results with an input of 6, it is disconcerting to see the strategy’s backtests fall apart with such a relatively minor shift. This is especially true given the fact that the NZDUSD and USDJPY pairs go from some of the strategy’s best performers at 6 days to the worst and second-worst losers on the minor shift in input.
Though not as dramatic, the strategy likewise drops off in performance with an input of 7 days. The chart below emphasizes that the result on an input of 6 days is so dramatically different than its nearby neighbors that it may have simply been a function of chance and not legitimate trading merit.
Backtested Automated Trading Results Seem Too Good to be True
Where does this leave us exactly? One could argue that the strategy does hold legitimate promise with an input of 6 trading days. Yet the actual period does not arguably make intuitive sense. What is a 6-day trading period and why does such an input so significantly outperform 5 days—a full trading week?
Given such doubts, we cannot reasonably advocate trading such a system. Though the equity curves look very impressive, the instability of results across optimization parameters suggests that such outperformance may be a fluke. Certainly there is one way to dispel skepticism: forward-testing the results and observing strategy performance. Yet the low frequency with which this strategy trades would require a very lengthy forward test to prove or disprove our convictions.
There are few absolutes in trading, and this is certainly not one of them. Yet there are arguably more promising automated trading strategies that warrant further attention before we dedicate months to a forward test of such a trading system. Other inputs for our strategy likewise show relatively meager risk-adjusted returns and do not seem to show much promise on backtests.
If you would like to suggest ideas for this topic or any other forex strategy you would like to see in this series, feel free to e-mail author David Rodríguez at firstname.lastname@example.org. To be added to this author’s distribution list, e-mail with subject line “distribution list”
View previous articles in this series:
Written by David Rodríguez, Quantitative Strategist for DailyFX.com