Technical analysis, or the statistical analysis of past price changes with the objective of forecasting future price changes, has been a hotly debated topic met with skepticism in many financial circles. Most traders and investors fall into one of three camps: Those who believe it's a science that works, those insisting that it's a self-fulfilling prophecy and those convinced it's worthless as a tool of prediction.
TUTORIAL: Technical Analysis: Introduction
Today's technical analysis is a far cry from what existed in the past. The development of neural networks, genetic algorithms and similar technologies, has dramatically improved accuracy in predicts and may mark a shift in the industry. In this article, we'll take a look at some empirical evidence to finally put this question to rest, with a specific focus on the foreign exchange (forex) market. (To know more about genetic algorithms, read: Using Genetic Algorithms To Forecast Financial Markets.)
Does Technical Analysis Really Work?
Researchers have been skeptical of technical analysis ever since Eugene Fama and Marshall Blume found buy-and-hold preferable to any filtering techniques in "Filter Rules and Stock-Market Trading." However, research focused on foreign exchange markets has shown unusually large profits driven by technical analysis, that have challenged the efficient market hypothesis.
In 1995, Blake LeBaron published a study called "Technical Trading Rule Profitability and Foreign Exchange Intervention," which proposed a possible reason why technical analysis was so effective in foreign exchange markets. The report found that predictability is largely reduced, if not eliminated, when discounting days that the Federal Reserve was actively intervening.
The underlying reason for the effectiveness of technical analysis in foreign exchange markets may, therefore, be that the priorities differ between major players. Unlike unpredictable equity markets, central banks have a strong incentive to keep currency prices at certain levels, which may make price movements more predictable, especially when they intervene. (To gain some knowledge about technical analysis practitioner, read: The Pioneers Of Technical Analysis.)
Neural Networks and Technical Analysis
With their ability to identify obscure patterns in data, neural network models have been growing in popularity. The models can approximate any nonlinear function to an arbitrary degree of accuracy, which makes them ideal for forecasting in many different settings. Moreover, modern software has made these networks accessible to even individual traders and investors.
Recent studies have been focused on using neural networks to identify underlying technical trading rules. In "A Case Study on Using Neural Networks to Perform Technical Forecasting of Forex," Jingtao Yao and Chew Lim Tan found that buy-and-hold strategies may be better than trend-following, but neural network models outperformed both, even when using just simple indicators like moving averages.
Another study called "Using Recurrent Neural Networks to Forecasting of Forex," provides more empirical evidence that neural networks can provide statistically reliable prediction of foreign exchange rates. The model used in the study reportedly obtained 80% accuracy in prediction, confirming that neural networks can be very effective in making foreign exchange predictions. (To know more about neural networks, check out: Neural Networks: Forecasting Profits.)
Components of an Effective System
There are several key considerations to take into account when developing a technical analysis-based trading system for the foreign exchange markets, according to the aforementioned studies on the topic. Here are a few key points to keep in mind when developing a system:
- Stick to the Swiss Franc and Japanese Yen. Several studies have found that the CHF and JPY are the two currencies that are easiest to predict. The prevailing theory behind this phenomenon appears to be the fact that these currencies are most prone to intervention, which is likely because they are both safe-haven currencies for international investors.
- Use Neural Networks to Optimize Systems. Neural networks have the ability to identify obscure patterns in data, which makes them perfect for foreign exchange markets. As a result, most of the current research on the subject centers around neural networks.
- Moving Averages and Logarithmic Returns. At least one study suggested that moving averages and logarithmic returns are the two best inputs for foreign exchange trading models, particularly when analyzing CHF or JPY.
A Word from the Opposition
The effectiveness of technical analysis-based trading systems is still disputed by many researchers. By using suspect testing data samples or excessively-customized trading systems, these researchers believe that the results from these studies may be misleading. Ultimately, this is difficult to tell without applying the system to new data sets, but traders should be aware of the concerns.
The two key concerns include:
- Data Dredging. Some studies may have used data mining techniques to identify misleading relationships in data. In this case, the performance of a test system may be valid within its testing data, but it wouldn't have any statistical significance in a wider population sample.
- Curve Fitting. Some studies may have used curve fitting techniques that may produce reliable results for one data set, but again, not for a wider population sample.
The Bottom Line
Technical analysis may not be proven to work in the equity markets, but there is growing evidence of its efficiency in the forex markets. The success has largely been attributed to the predictability of interventions seen largely in safe-haven currencies by central banks. However, some researchers remain skeptical, given the potential for at least two types of misleading analysis techniques.