The views expressed are those of the individual authors and do not necessarily reflect official positions of the Federal Reserve Bank of St. Louis, the Federal Reserve System, or the Board of Governors. Federal Reserve Bank of St. Louis Working Papers are preliminary materials circulated to stimulate discussion and critical comment. References in publications to Federal Reserve Bank of St. Louis Working Papers (other than an acknowledgment that the writer has had access to unpublished material) should be cleared with the author or authors. Photo courtesy of The Gateway Arch, St. Louis, MO. www.gatewayarch.com

This paper presents results from an experimental computer simulated stock market. In this market artificial intelligence algorithms take on the role of traders. They make predictions about the future, and buy and sell stock as indicated by their expectations of future risk and return. Prices are set endogenously to clear the market. Time series from this market are analyzed from the standpoint of well-known empirical features in real markets. The simulated market is able to replicate several of these phenomenon, including fundamental and technical predictability, volatility persistence, and leptokurtosis. Moreover, agent behavior is shown to be consistent with these features, in that they condition on the variables that are found to be significant in the time series tests. Agents are also able to collectively learn a homogeneous rational expectations equilibrium for certain parameters giving both time series and individual forecast values

Numerous studies in the finance literature have investigated technical analysis to determine its validity as an investment tool. Several of these studies conclude that technical analysis does have merit, however, it is noted that the effects of data-snooping are not fully accounted for. In this paper we utilize White's Reality Check bootstrap methodology (White (1997)) to evaluate simple technical trading rules while quantifying the data-snooping bias and fully adjusting for its effect in the context of the full universe from which the trading rules were drawn. Hence, for the first time, the paper presents a means of calculating a comprehensive test of performance across all trading rules. In particular, we consider the study of Brock, Lakonishok, and LeBaron (1992), expand their universe of 26 trading rules, apply the rules to 100 years of daily data on the Dow Jones Industrial Average, and determine the effects of data-snooping. During the sample period inspected by Brock, Lakonishok and LeBaron, we find that the best technical trading rule is capable of generating superior performance even after accounting for data- snooping. However, we also find that the best technical trading rule does not provide superior performance when used to trade in the subsequent 10-year post-sample period.

Recent research has discovered two seemingly contradictory facts about U.S. intervention in foreign exchange markets. On the one hand, extrapolative technical trading rules trade against U.S. foreign exchange intervention and produce excess returns—returns in excess of nominal interest rates—during these periods, and U.S. intervention itself, is profitable over long periods. LeBaron (1996) and Szakmary and Mathur (1997) have shown that excess returns to technical trading rules are high during periods of central bank intervention and that the technical rules trade contrary to the direction of official intervention. Along the same lines, Neely and Weller (1997) have shown that trading rules constructed by genetic programs can use information on the direction of U.S. intervention to increase their excess returns in some exchange rates: When the Federal Reserve is buying dollars, traders following technical rules are usually selling dollars and profiting handsomely. Some—Dooley and Shafer (1983), Corrado and Taylor (1986), Sweeney

This paper examines the efficiency of the forward yen/dollar market using micro survey data. Conventional tests of unbiasedness do not correspond directly to the zero-profit condition. Instead, we use the survey data to calculate potential profits of individual forecasters based on a natural trading rule. We find that although the survey data are not the best predictor of future spot rates in terms of typical mean square forecast error criteria, the survey data can be used to obtain on average positive profits. However, these profits are small and highly variable. Similar results are found when we examine profits generated by a trading rule using regression forecasts. The profits are found to be correlated with risk type variables but not other available information. Key Words: Foreign exchange rate; Expectations; Forward rate; and Efficient markets. JEL classification: F31, G14, G15 Acknowledgment: We thank J. Frankel, A. Timmermann, A. Melino, an anonymous referee and seminar partici...

• Among the technical trading signals supplied to customers by foreign exchange trading firms are “support ” and “resistance ” levels. These levels indicate points at which an exchange rate trend is likely to be interrupted or reversed. • A rigorous test of the levels specified by six trading firms during the 1996-98 period reveals that these signals were quite successful in predicting intraday trend interruptions. • Although all six firms were able to identify turning points in exchange rate trends, some firms performed markedly better than others. As a group, the firms predicted turning points in the dollar-yen and dollar-pound exchange rates more accurately than turning points in the dollar-mark exchange rate. • In addition, the predictive power of the support and resistance levels appeared to last at least five business days after they were first communicated to customers. arly in the morning of each business day, the major foreign exchange trading firms send their customers lists of technical trading signals for that day. Timely technical signals are also supplied by major real-time information providers. These signals, which are based primarily on prior price and volume movements, are widely used by active foreign exchange market participants for speculation and for timing their nonspeculative currency transactions. In fact, 25 to 30 percent of foreign exchange traders base most of their trades on

This paper tests whether fitted linear models can replicate results from moment tests inspired by moving average technical trading rules for weekly foreign exchange series. Estimation is performed using standard OLS and maximum likelihood methods, along with a simulated method of moments technique which incorporates the trading rule moments into the estimation procedure. Results show that linear models are capable of replicating the trading rule moments along with the small autocorrelations observed in these series. This result holds for parameter values estimated using SMM and GARCH disturbances, but not for parameters estimated using maximum likelihood. The estimated models are simulated to examine the amount of predictability over long horizons.

This article proposes the use of recurrent neural networks in order to forecast foreign exchange rates. Artificial neural networks have proven to be efficient and profitable in forecasting financial time series. In particular, recurrent networks, in which activity patterns pass through the network more than once before they generate an output pattern, can learn extremely complex temporal sequences. Three recurrent architectures are compared in terms of prediction accuracy of futures forecast for Deutsche mark currency. A trading strategy is then devised and optimized. The profitability of the trading strategy, taking into account transaction costs, is shown for the different architectures. The methods described here, which have obtained promising results in real-time trading, are applicable to other markets. For years, opposing views have existed between the trading and academic communities about the statistical properties of foreign exchange rates. Traders considered exchange rates to have persistent trends that permitted mechanical trading systems (systematic methods for repeatedly buying and selling on the basis of past prices and technical indicators) to consistently generate profits with relatively low risk. Researchers, on the other hand, presented evidence supporting the random

Previous studies have reported mixed results regarding the success of technical trading rules in currency markets. Abnormal returns were observed in many studies using data up to the mid 1980s, while more recent studies generally report less success for technical trading rules. This paper tests whether moving average trading rule profits have declined over the period from 1971-2000. If so, previous profits may represent a temporary inefficiency that has since been eliminated in the currency markets. The hypothesis is tested using 18 exchange rate series over a longer time period than in previous studies. Rules are optimized for successive 5-year in-sample periods from 1971-95 and tested over subsequent 5-year out-of-sample periods. Results show that risk-adjusted trading rule profits have declined over time—from an average of 3.5 % in the 1970s to about zero in the 1990s. Thus, market inefficiencies reported in previous studies may have been only temporary inefficiencies. 2

and the participants of the Taipei conference on “Analysis of High-Frequency Financial Data and Market Microstructure ” for their valuable comments and suggestions. We also thank P. R. Hansen for sharing his