One of the great challenges for participants in FX has been how to more effectively manage their positions in this highly unpredictable market. Typically, FX markets behave in one of two ways: trending, where currencies move in one steady direction over a period; or ranging, where currencies move within a range.
Traditionally, currency managers have used one of two strategies: trend following, which takes positions on long-term trends; and volatility capture, which uses option positions to sell volatility between trends.
Adopting a single strategy can cause losses. A market that is ranging undermines maintained long or short positions, while short volatility positions fail to take advantage of trending movements in the spot rate over the long term.
One way to hedge this risk has been to adopt both strategies side-by-side. This limits the potential losses of a single strategy, however, profits are also limited.
Switching between trend following and volatility capture strategies is optimal if the trading regime can be predicted in advance. The ability to predict the type of trading regime in the coming period allows currency managers to benefit from most of the movement in major trends while also generating returns during range periods.
ABN AMRO has developed an FX model that indicates whether the EUR/USD exchange rate is likely to trend higher or lower, or be subject to movements within a trading range.
In back-tests, the model generated absolute and risk-adjusted returns above those of traditional trading strategies.
Despite the existence of strong longer-term trends, extended periods of range trading have been a striking feature of spot action in EUR/USD. In 2004, the prolonged period of range trading that lasted from approximately May until late September made for very difficult trading conditions for trend following strategies.
The model, dubbed Regime Prediction Indicator, is based on three factors: implied volatilities; speculative positioning data; and the strength of the trend. All data is timely, available at the moment of the allocation decision. ABN AMRO calculates the index using weekly data as of Wednesday's U.S. close and predicts the regime--trend or range--over the coming week.
There are two aspects to the index. The first is a 'switching' indicator, which recommends which strategy traders should use in the coming week. The second is a probability weighting index, which indicates how likely EUR/USD will be trending or trading within a range.
Switching Indicator
The approach is based on an all-or-nothing style whereby the entire portfolio is switched from one style to another, depending on the prediction of the index. This approach does not take into account the strength of the signal. Whichever style is forecast to perform better in the subsequent week is followed.
Probability Weighting
This approach adjusts the proportion of capital assigned to each trading style according to the probability of that style outperforming in the coming week. The probability of trend and the probability of range always add up to 100%.
Table 1 summarizes back-testing results and illustrates the benefits of using the RPI as opposed to traditional trading strategies. The results are discussed in detail at the end of the article. Since January 2003, the switching indicator has generated the highest absolute and risk adjusted returns. The probability weighted approach has the lowest volatility of returns, something which may be a priority for some currency managers.
The index was developed using historical data from 2000 through 2002 and applied out-of-sample from 2003 through Q1 of this year. The result is an increase in risk-adjusted returns from those of a basic trend-following strategy. In its switching formulation, the RPI leads to more than a doubling of raw returns relative to a basic trend-following model. In its probability-weighted formulation, it improves the information ratio by a factor of eight relative to a passive 50:50 blend.
Signals from the index are published each Wednesday after the U.S. close.
| Table 1 : | |||||
| Trading Rule Comparisons: | |||||
| Returns Based On Period Jan 2003 – March 2005. | |||||
| Annualized | Annualized | ||||
| mean | standard | Biggest | Hit Ratio | ||
| weekly | Information | deviation of | 1-week | (% winning | |
| Trading Rule | return | Ratio | weekly returns | loss | trades) |
| Trend following | 3.80% | 0.39 | 9.63% | -3.85% | 60.30% |
| Volatility capture | -2.16% | -0.38 | 5.70% | -2.78% | 51.70% |
| Static 50:50 | 0.57% | 0.09 | 6.17% | -3.31% | 69.80% |
| Switching | 8.41% | 1.08 | 7.77% | -3.85% | 61.20% |
| Probability weighted | 4.13% | 0.74 | 5.59% | -2.94% | 68.10% |
| Source: ABN AMRO |
Variables
Trend following: We use a simple five and 40-day moving average (MA) cross as our trend following rule. When the five-day MA is above the 40-day MA, the rule is long and vice-versa. For simplicity, we allow for flips from long to short only on Wednesdays and we employ no trailing stop. This strategy has a long history of profitability. From the launch of the euro at the start of 1999 through March 30 this year, the annualized average weekly return is 5.28% and the information ratio is 0.51. In the out-of-sample period between 2003 and March this year, the trend rule is profitable in 60% of the 117 weeks and yields an average annualised weekly return of 3.80%.
Volatility capture (range trading): In our simulations we have simply sold a one-week at-the-money forward straddle at every Wednesday's close. For simplicity, we do not allow any delta hedging or other risk mitigation strategies. The profit is the payout less the premium collected. Since the launch of the euro through Q1 2005, the annualized average weekly return of this strategy is -2.75%. The rule worked in 52% of the 117 weeks in the out of sample period.
Passive diversification (50/50): In this approach we simply allocate 50% of our notional capital each to the trend following and volatility capture rules and run both concurrently. The average annualised weekly return from this approach for the life of the euro is 0.57% and the information ratio 0.09. In our out-of-sample period, the return is identical at 0.57% as is the information ratio at 0.09.
Switching Indicator (all or nothing): This is an indicator that uses data available on Wednesday's U.S. close to forecast which of the trend following or volatility capture strategies are likely to perform better in the upcoming week. Over our out-of-sample window, the Switching Indicator yields the highest mean annualised weekly return of 8.41%. It also yields the highest risk adjusted returns with an information ratio of 1.08.
Probability Index (weighting): The probability indicator is a forecast of the probability that the upcoming week will be a week in which the trend-following strategy is superior. We use it to allocate capital to the two rules. If the indicator gives a 75% probability to a trend and a 25% probability to a range in the upcoming week, then we allocate 75% of the capital to the trend-following strategy and 25% to the volatility capture strategy. The decision for the coming week is taken at the Wednesday U.S. close. The average annualized return from this approach is 4.13%, which is lower than for the switching approach or the trend following strategy taken alone. However, the big advantage of this approach is that it has the lowest volatility of returns with a standard deviation of weekly returns of 5.59%. The information ratio is twice that of trend following strategy taken alone. Because it is a more diversified approach than using the switching indicator or a simple trend-following approach, it would likely be the choice of currency managers seeking to minimize risk and maximize their hit ratio.
This week's Learning Curve was written by Aziz McMahon, FX strategist at ABN AMROin New York and Greg Anderson, FX strategist at ABN AMROin Chicago.
