Using Trading Mechanisms to Investigate Large Futures Data and Their Implications to Market Trends
Total Page:16
File Type:pdf, Size:1020Kb
Soft Comput DOI 10.1007/s00500-016-2162-6 FOCUS Using trading mechanisms to investigate large futures data and their implications to market trends Mu-En Wu1 · Chia-Hung Wang2 · Wei-Ho Chung3 © Springer-Verlag Berlin Heidelberg 2016 Abstract Market trends have been one of the highly debated the existence of the momentum effect via applying these two phenomena in the financial industries and academia. Prior new trading strategies. Besides, we analyze the market trends works show the profitability in exploiting transactions via through the repeated simulations of random trades with the market trend quantification; on the other hand, traders’ stop-loss and stop-profit mechanisms. Our numerical results behaviors and effects on the market trends can be better reveal that there exist momentum effects in TAIEX Futures, understood by market trend studies. In general, the trading which verifies the market inefficiency and the market prof- strategies on the market trend include trend following strate- itability in exploiting the market inefficiency. In addition, gies and contrarian strategies. Following the trend, trading the techniques of random trades are also applied to the other strategies exploit the momentum effects. The momentum commodities, such as AAPL in NASDAQ, IBM, GOOG in strategies profit in a long position with the rising market NYSE, and, TSMC in TPE, and so on. Surprisingly, not all the prices, as well as in a short position with the decreasing mar- stocks have the momentum effects. Our experimental results ket prices. On the contrary, the view of contrarian trading show that some stocks or markets are more suitable for the strategy is based on the mean-reversion property, i.e., a long mean-reverse strategy. Finally, we propose a technique to position is taken when the price moves down and a short posi- quantify the momentum effect of a financial market by using tion is taken when the price moves up. In this paper, we apply Jensen–Shannon divergence. the stop-loss and stop-profit mechanisms to verify the market trends based on two new simple strategies, i.e., the BuyOp. strategy and the BuyHi.SellLo. strategy. We back-test these 1 Introduction two strategies on the Taiwan Stock Exchange Capitalization Weighted Stock Index Futures (TAIEX Futures) during the The trend of financial prices has been one of the most con- period from May 25, 2010 to August 19, 2015. We compare cerned phenomena for many investors and in academia. the numerical results of its profits and losses through vari- Researchers develop profitable trading strategies through ous stop-loss thresholds and stop-profit thresholds, and verify exploiting market behaviors, technical analysis (Park and Irwin 2007; Schulmeister 2009; Holmberg et al. 2012; Borda Communicated by C.-H. Chen. et al. 2011; Cekirdekci 2010) and commodity pricing model, etc. However, the market is often conjectured to be unpre- B Wei-Ho Chung dictable, which conforms to the observation where only [email protected] few transaction strategies are profitable in practice. Besides, losses in transactions are usual for most investors in finan- 1 Department of Mathematics, Soochow University, Taipei, Taiwan cial markets. Hence, in this paper, we investigate the price behavior and market trends. 2 College of Information Science and Engineering, Fujian University of Technology, Fuzhou, Fujian, One model to describe the price behavior in stock markets People’s Republic of China is to use the random walk to explain unpredictability (Basu 3 Research Center for Information Technology Innovation, 1977; Brown 1971). In 1970, Fama (1970) proposed the Academia Sinica, Taipei, Taiwan efficient-market hypothesis (EMH, Fama 1970), which con- 123 M.-E. Wu et al. sists of three types: strong form, semi-strong form, and weak In this paper, we investigate the profitability of trading form. These three forms differ in whether or not the public strategies in the studied market, and the impact of the stop- information, non-public information, or historical informa- loss and stop-profit mechanisms on the profit and loss to ver- tion have been reflected in the fluctuations of the stock price. ify whether there exist the momentum effects in the studied Even though EMH (Fama 1970) has been studied widely, financial market. Here, the Taiwan Stock Exchange Capi- there are many evidences reveling that certain financial mar- talization Weighted Stock Index Futures (TAIEX Futures) kets do not conform to market efficiency, such as Hung et al. is used as the case study for the simplicity. Other financial (2014). On the other hand, other research works show the markets can be studied via a similar approach; this can be market efficiency in some financial markets (Brown 1971; considered for a future work. We back-test the real intra-day Hung et al. 2014; Fama 1970), where the strategies are unable data during the period between May 25, 2010 and August to make consistent profits in certain financial markets. How- 19, 2015, by 1-min time frame data, including the opening ever, many research works show the market may not be price (Op.), closing price, the highest price (Hi.), and the efficient in the stock markets (Holmberg et al. 2012; Borda lowest price (Lo.). In the numerical experiments, day trading et al. 2011; Cekirdekci 2010; Ansari and Khan 2012; Tsai strategy is adopted. First, a simple strategy is used, BuyOp.: et al. 2014). Park and Irwin (2007) made a survey and found long, a position at the open price for daily trading; and close, a rapid increase in the amount of literature in studying tech- the position at the end of the market, to observe the changes nical analysis. Schulmeister (2009) studied the momentum in the profit–loss curve through back-testing various stop- and reversal effects in the S&P 500 spot and futures mar- loss points and stop-profit points. Secondly, another simple ket by using technical trading systems. These works show strategy used, BuyHi.SellLo., is conceptually more aligned the existence of market inefficiency and its resulting prof- with the momentum effects. BuyHi.SellLo. is to long\short itability. In the following, we introduce two types of trading a position at open if the open price is higher\lower than the strategies: momentum trading strategy and mean-reversion close price of the previous trading day, and then close the strategy. positions at the end of the market for the daily trading. To The trend following trading strategy assumes that the avoid influences of a specific strategy on the profit and loss, movement of market price is driven by the momentum, i.e., we propose the concept of random trades to test the exis- the market price moves in the direction of the momentum. tence of a momentum effect in the market. The technique Typically, the characteristic of momentum trading strategy of random trade is to BUY or SELL at daily open price with is to long a position upon the price rising up to a prede- 50 % probability. The experiments are conducted by simulat- fined price threshold. On the contrary, the mean-reversion ing 10,000 rounds to observe the profit–loss distribution via strategy is to shor a position upon the price dropping below a different stop-loss and stop-profit points. If the momentum predefined price threshold. The trend following trading strat- effect exists in the market, the profit distribution with stop- egy is based on the momentum effect proposed by Jegadeesh loss and no stop-profit should smoothly move to the profitable and Titman (1993). The momentum effect represents that the side. This phenomenon also appears on the other markets, prices of commodities will continue the trend in its direction such as NASDAQ, NYSE, TPE, TYO, and so on. Finally, of movement, i.e., the trend of the commodity price will con- we use the method of random trades to define an index of tinue in time. Based on the momentum effect, it is possible momentum effect, which is expected to serve as one of the to develop profitable momentum trading strategies through standard measures of momentum effect for general financial buying stocks with increasing prices and selling decreasing markets. Note that the money managements (Tharp 2008; stocks. There are many aspects on momentum trading strate- Vince 2012; Vince and Zhu 2013b; MacLean and Ziemba gies, one of which is the effect of fat tails. Many studies have 2006) are also crucial in the investment outcomes. To avoid shown that there exist fat tails in the financial markets (Park ambiguity and simplify the interpretation of the numerical and Irwin 2007; Borda et al. 2011; Tsai et al. 2014). The results, the money management technique is not adopted in chances of rally or crash often exceed the level of belief in our back-testing strategies. The purpose of this work is to ver- most investors. ify and classify the market trends via well-calibrated trading On the other hand, the contrarian trading strategy is the strategies. The technique of money management (Zhu et al. reverse of the momentum trading strategy. The contrarian 2012; Vince and Zhu 2013a; Zhu 2007; de Prado et al. 2013) trading strategy is based on the mean-reversion theory. That can be considered as future works to be applied to the basic is, if the current price continues to increase and is already strategies. higher than the average price above a threshold, there is a The organization of this paper is shown as follows. In Sect. great chance that the price will fall back to the average. Sim- 2, we show the preliminaries in this work, including the two ilarly, if the current price continues to decline and is lower types of trading strategies, and the financial market for the than the average below a threshold, there is a great probability back-testing and the device for the experiments.