Avoiding forex disaster through non-correlation

March 31, 2011 07:00 PM

An ever-present danger facing spot forex traders is the constant focus on short-term opportunities while missing the big picture.

Most recently, traders have witnessed the risk created by major geopolitical events. An exogenous event such as the Libyan revolt can cause a wave of risk aversion that cascades throughout the markets. If the best and the brightest in the world can’t predict these events, nor the best trading systems, the forex trader needs to develop a defensive or risk-reduction strategy and minimize the risk of being swept away on the wrong side of such events. While we cannot predict unexpected events, we can choose to lower the risk in spot forex trading by selecting non-correlated positions.

How does the EUR/USD correlate on a one-week, one-month or yearly basis with other currency pairs? Are there surprises? Using correlation data, we can look for anomalies or outlier patterns in current currency correlations. For example, conventional wisdom regarding the EUR/USD and the USD/CHF pairs is that they are about 90% negatively correlated. That, however, is belied by the contemporary correlations data (see "The forex matrix"), and allows for the possibility of a reversion to the mean trade.


But the important point to remember with correlations is to have as many non-correlated positions on as possible to avoid huge drawdowns in big moves. If your models like the euro and pound — two highly correlated pairs — against the dollar, you need to select the stronger signal and go with it rather than doubling your exposure by taking both trades. Then you only should add trades with non- or negatively-correlated pairs.

For traders putting on multiple currency pairs, finding the most diverse set of currency pairs provides extra protection. The familiar adage of don’t put all your eggs in one basket applies here. If a trader wanted to apply this principle of minimum correlations and trade multiple currencies, the first step is to select a prime currency pair and then choose other pairs that have low correlations. For example, if we selected the EUR/USD, the set of low correlated currency pairs alongside it would include the USD/MXN and the USD/CAD. If we selected the GBP/USD as the main pair, then a potential set of low-correlated pairs would be the USD/MXN and EUR/AUD (see images below).


It’s worthwhile to look at the AUD/USD from a correlation point of view. An AUD/USD low correlation basket includes the EUR/JPY, EUR/GBP or EUR/CAD. The AUD/USD and the EUR/CAD have a one-month lowest correlation of -0.03%. A pattern that traders need to be aware of is the fact that when you add more currency pairs, you lose non-correlation.

We haven’t talked about signals and we should be clear that these are not arbitrage trades, but trades based on their own merit that we, through a study of correlation, are ensuring are not adding risk. The correlation between or among each position you carry needs to be understood in order to understand your total risk exposure.

There are significant strategic trading implications from a correlation analysis for beginners as well as more experienced traders. First, while the EUR/USD is very popular, it really provides the greatest risk exposure to trade along with other currencies. Because its correlations are very high with other pairs, losses from trading the EUR/USD would be more likely replicated in the other pairs. The currency pair that provides the best non-correlation with other pairs is the AUD/USD. For beginning and even more experienced traders trading the AUD/USD with the EUR/JPY or the EUR/GBP, the USD/MXN and USD/CAD offers a basket that generates the greatest potential to stabilize performance.

Abe Cofnas is the author of "Sentiment Indicators" (Bloomberg Press). He can be reached at abecofnas@gmail.com.

About the Author

Abe Cofnas is author of “Sentiment Indicators” and “Trading Binary Options: Strategies and Tactics” (Bloomberg Press). He is editor of binarydimensions.com newsletter and can be reached at abecofnas@gmail.com.