Nevertheless, the behaviour of the technical trading rules based on the learning methods was the complete opposite during the falls subperiod (2 September 2000 to 31 December 2002). Thus all out-of-sample learning methods overcame the return of the B&H strategy, especially the fi ltered Boosting model. Also, the Sharpe ratio of the fi ltered Boosting model was higher than that of B&H, and much higher than the Sharpe ratio of the Committee and Bayesian models. The same occurred with the ideal profi t ratio.
Also, it was observed that, except during the period of generalized rises in the market, the fi ltered Boosting model obtained more net return than any moving average model, supporting less risk that any other.
Therefore, although the fi ltered Boosting model is not able to overcome the returns of the B&H strategy during the rising period, results obtained suggest that it does overcome the B&H during the falling period and is able to absorb a large part of falls in the market. Therefore, the fi ltered Boosting model could be used as a conservative strategy devoted to diminishing risk during eventual market collapses.
Nevertheless, it is necessary to be cautious and study, in future research, how robust the fi ndings in this paper are to other fi nancial indexes, and it could be helpful to investigate other popular indexes (equity, bonds or foreign exchanges).
The fi nal conclusion of this research is not to cast doubt upon the predictive power of moving average rules in the series and period analysed. On the contrary, our results support that although the moving averages show contingency and variability of its predictive power, a wide set of moving averages has more predictive information than any individual one, from which the fi ltered Boosting algorithm may take advantage.
This research points out that the use of learning methods like boosting seems more robust and profi table than the use of individual moving averages. It calls attention to technical analysts for using Boosting and other combining prediction methods instead of individual moving averages.
Finally, this research could be extended to a broader set of technical trading rules like those used in Sullivan et al. (1999); that is, fi lter rules, support and resistance, channel break-outs and onbalance volume averages.
This research is supported by the Spanish Ministry of Science and Technology through the project SEJ2006-07701 ELON.
Alexander S. 1961. Price movements in speculative markets: trends or random walks. Industrial Management Review 2: 7–26.
Allen F, Karjalainen R. 1999. Using genetic algorithms to fi nd technical trading rules. Journal of Financial Economics 51: 245–271.
Blume L, Easley D, O’Hara M. 1994. Market statistics and technical analysis: the role of volume. Journal of Finance 49: 153–181.
Bodie Z, Kane A, Marcus AJ. 2002. Essentials of Investments. MacGraw-Hill: New York.
Brock W, Lakonish J, LeBaron B. 1992. Simple technical rules and the stochastic properties of stock returns. Journal of Finance 47: 1731–1764.
Brown D, Jennings R. 1989. On technical analysis. Review of Financial Studies 4: 527–551.
Chang PHK, Osler CL. 1999. Methodical madness: technical analysis and the irrationality of exchange-rate forecasts. Economic Journal 109: 636–661.
Fama EF. 1970. Effi cient capital market: a review of theory and empirical work. Journal of Finance 25: 383– 417.
Fama EF. 1976. Foundations of Finance. Basic Books: New York.
Fama EF, Blume ME. 1966. Filter rules and stock market trading. Journal of Business 1: 226–241.
Freund Y, Schapire RE. 1997. A decision-theoretic generation of on-line learning and application to boosting. Journal of Computer and System Sciences 55(1): 119–139.
Уважаемый посетитель!
Чтобы распечатать файл, скачайте его (в формате Word).
Ссылка на скачивание - внизу страницы.