Empirical study of trading rule discovery in China stock market

https://doi.org/10.1016/j.eswa.2004.12.014Get rights and content

Abstract

This case study employs classic knowledge engineering methods. We found a new template grid from Shanghai Stock Exchange, and compared it with the classic one. The results show that the new template grid is more effective than the classic one. Then, we built a trading rule named Rule 1v. The finding of results using our trading rule, which is significantly better than the overall average for the period, constitute a failure to confirm our implicit null hypothesis, which is the efficient market hypothesis.

Introduction

Charting, a technique of technical analysis, compares market price and volume history to archetypal chart patterns and predicts future price behavior based on the degree of the match. The charting technique may be operationalized through trading rules of the form:

If charting pattern X is identified in the previous N trading days, then buy; and sell on the Yth trading day after that. If charting pattern X is identified in the previous N trading days, then sell.

Obviously, the difficult part is to identify ‘charting pattern X’. With a good ‘charting pattern’, we can find valid trading rules. Charting is rarely tested in the academic literature, but it is a good technique to find new knowledge from time series data. Martinelli and Hyman (1998) and Leigh, Modani, Purvis, and Roberts (2002) imply that trading success may be achieved with charting. Nobel prize winner Paul Samuelson (1965, p. 44) had the following to say in regard to this subject matter: “…there is no way of making an expected profit by extrapolating past changes in the futures price, by chart or any esoteric devices of magic or mathematics. The market quotation already contains in itself all that can be known about the future, and in that sense it has discounted future contingencies as much as is humanly possible.”

There are two stock exchanges in the mainland of China, which is Shanghai Stock Exchange and Shenzhen. Shanghai Stock Exchange is the most popular one in China, so we therefore used the Shanghai Stock Exchange Composite Index. In this paper, the closing price values for Shanghai Stock Exchange Composite Index for 2808 trading days in the period, January 4, 1993–June 30, 2004 are employed to find a new charting, and set up a rational trading rule. At the same time, the charting pattern provided by Leigh, Paz, and Purvis (2002) is validated. The finding of results using our trading rule which is significantly better than the overall average for the period, constitute a failure to confirm our implicit null hypothesis, which is the efficient markets hypothesis.

Section snippets

Methodology

Charting, the sort of technical analysis that we used, is based on the recognition of certain graphical patterns in price and/or volume time series data. This work concentrates on one charting pattern, the bull flag. Downes and Goodman (1998) defined a flag as a technical chart pattern resembling a flag-shaped-like parallelogram with masts on either side, showing a consolidation within a trend. It results from price fluctuations within a narrow range, both preceded and followed by sharp rises

Results

From Table 3, we can see that T2 performs better, and the horizon interval of 40 trading days has a higher profitability than the other interval. The highest FITratio for T2 is 92.5% on January 10th 1996, which is a beginning of a big bull market. The results in Table 3 reveal that excess profits might be realized through a trading rule using this approach. Trading rule, which might be derived from these data mining results includes:

Rule 1. If template grid T2 is identified in the previous 60

Conclusion and implication

The T-test probability for Rule 1v is much higher, which means that there is not a significant difference between the market average returns and the trading rule average returns. Can we still accept the result of Rule 1v? To answer this question, we must review the process of T-test in pair. With a higher standard deviation, the T-test probability will be higher, and from the process of trading rule average returns, we can see that the trading rule average return for overall means the

Acknowledgements

The authors are grateful to the 863 (2003AA413033 and 2002AA414060-7) who sponsored this research.

References (15)

There are more references available in the full text version of this article.

Cited by (0)

View full text