The Fed’s future path still seems more bullish than the European Central Bank. If so, the yiel...
Revisiting the Forex Holy Grail
03/27/2014 9:00 am EST
Adam Lemon of DailyForex.com details the pitfalls of following a statistical approach to trading the forex market.
Last week I wrote about the market phenomena that is the true “holy grail” of the forex markets. The phenomena is the statistical tendency of market to produce excessive returns, which can be “gamed” profitably by letting winning trades run to large reward to risk multiples, while cutting losing trades short by using relatively tight stop losses. If the most volatile instruments are traded in this style, it is possible to be nicely profitable over time without having to really make any analysis or decisions. Despite that, this path has some serious pitfalls that must be avoided intelligently. In this week’s article, I will go into more detail about what those pitfalls are, and some of the best ways to deal with them.
Back to the Data
We can begin by taking a look at the historical data showing how entries upon next bar breaks of H4 engulfing candles performed on the most volatile instruments from 2011 to 2013, a three-year period, depending upon the reward to risk multiples that might have been selected as targets for trade exits:
This table contains two immediately useful pieces of information. Firstly, we would have taken a total of 2,810 trades. Secondly, the positive expectancy per trade rises dramatically until a reward to risk ratio of 25:1 is reached, after which it rises very slowly before falling off a cliff at above 50:1. Let’s say we have been following the 25:1 model. This data is not shown in the above table, but of those 2,810 trades taken, only 139 were 25:1 winners. This means that approximately 95% of the trades were losing trades. These numbers would put a severe strain on any kind of money management strategy, as the probability of suffering enormous losing streaks would be extremely high. It is more likely than not there was a streak of between 100 and 120 consecutive losing trades during that three year period.
There are three possible ways to improve the methodology:
- Be more selective with entries
- Be more selective with exits
- Risk a consistent and very small percentage of capital per trade (money management)
Let’s address each one in turn.
Our problem is that we are currently set to enter a very large number of trades, the vast majority of which will be losers. If we can find a way to enter significantly less trades without suffering a proportionate fall in the expectancy per trade, we can worry less about the strain of likely losing streaks.
The danger here is that when profit rests upon a relatively small number of winning trades, you have to be very careful not to cut yourself out of many of those. Fortunately, using the historical data from 2011 to 2013, there seems to be a relatively simple filter which does the job.
To win large trades, a trend has to be present. In an uptrend, the price pulls back within the trend making a major low, and then resumes its original direction. By only taking engulfing candles in such an uptrend that make a low lower than the previous four candles, or that directly follow such a candle, we are able to filter out a lot of the losing trades, without sacrificing too many of the winning trades. Here is a table of the performance over the same three year period using this entry filter:
It can be seen that overall, the total number of trades is reduced by slightly more than one third, but the winning trades tend to be reduced by a smaller percentage, resulting in rises in the expectancies from 3:1 to 50:1. The probable consecutive losing streak is reduced to somewhere between 80 and 90 trades, which is also an improvement. It is noticeable that this filter had a strongly negative effect upon the gold trades.
NEXT PAGE: Selective Exits & Money Management|pagebreak|
Other entry filters that could improve performance would include entering only after engulfing candles with relatively small ranges, as the total positive distance required to be a winner is shorter. Time of day and trend filters can also be applied, although these can be pretty risky. For example, gold tends to short well before the London open and long well after the London close. The yen pairs tend to perform well following the first candle representing the initial few hours of the Tokyo session. Bounces off major support or resistance levels can also be the origins of good trades, although it is surprising how many of the best resumptions within trends begin ahead of these levels.
So far, we have only looked a methodology that exits at a fixed R multiple. This could be refined by setting a target based upon an average volatility or number of pips, so that trades with larger risks can be exited at smaller R multiples. Additionally, there is the question of raising stop losses to break even and beyond. We have no hard data, but it is likely that moving the stop loss to break even somewhere between two days and one week after entry, or after the trade has moved a certain favorable distance, would enhance the results. Caution is required here as there are often retests of entry zones in long-term position trading using an H4 chart.
Of course, should the instrument being traded in an uptrend fail to make a major higher high a little way short of the desired target, it would make sense to exit at that point and take the profit.
It is vital to use robust and intelligent money management techniques to minimize the risk of catastrophic loss. As a losing streak of 80 consecutive trades was probably during the three-year period, risking a percentage of capital rather than an absolute amount based upon the starting capital is essential. For example, risking 1% of the starting capital on each trade would result in an 80% loss at the end of the losing streak followed by a 25% addition by the first winning trade, resulting in a total of 45%. Risking 1% of the total capital would result in a 55% loss, followed by a recovery of approximately 17%, resulting in a total of 72%.
Always bear in mind that the more of your account you lose, the harder it becomes to make it back. A 50% loss requires a 100% gain just to get back to break even.
A more appropriate risk per trade would be something like 0.25% of the account, which would result in a draw down to about 81% of the starting total after the likely losing streak, recovering to about 93% after the close of the first winning trade. It all depends upon your individual risk tolerance and tolerance of account draw down.
One final warning: when you are trading correlated pairs, as in this example where three of the four instruments are yen crosses, an additional defensive measure can be taken of reducing the total risk when taking multiple trades at the same time in the same direction. This will be especially important where all the trades are long yen. In fact, a careful study might show that the best trades are the ones that set up on all three of the yen pairs simultaneously, or at least that this situation produces an enhanced statistical edge.
Trading in this systematic way requires careful study of historical data, without curve fitting. Before trying this with real money, test rigorously and be honest in answering your own questions, and be sure to study thoroughly and carefully.
By Adam Lemon, Contributor, DailyForex.com
Related Articles on FOREX
Trade idea: No guarantees here of course, but maybe it’s a small caution flag for dollar bulls...
As of August 2015, renminbi (RMB) in payments globally accounted for 2.8 percent of the total, the f...
Our favorite horse to ride here for a “correction” lower would be the euro. And we would...