Research On Program Trading Based On The Trend Reverse

The program trading investment is realized by computer through the logic operation, and expressed by computer language. Now it is widely used in domestic and foreign investment and transactions in financial market. In 2013, by an accidental opportunity, in a competition of the simulation of futures, I used "trend reversal" strategies and obtained very good results. This study is mainly about the program trading based on "trend reversal" strategy. It studies the realization of the program trading based on "trend reversal" strategy theoretically and empirically.Theoretically, we discuss the formation of EMH and the development of FMH, and then from the viewpoint of fractal, explain that the price trend is "biased random walk" by the use of R/S analysis(Rescaled Range Analysis). Empirically, we studied 10201 kinds of possible parameters, and based on the special combination, we studied the model validity. We also make the forecast of the model parameters. This paper is mainly divided into three parts:The first part explains the classical theories, i.e. the efficient market hypothesis and the market fractal hypothesis. The time series of securities prices is studied by the literature method. The trend is explained by the use of R/S method and the applicability of "trend reversal" strategy is analyzed theoretically.The second part introduces the realization of program-trading model based on the "trend reversal" strategy. We focus on the explanation of the logical expression of strategies, and the detailed description of the related parameters. By the study of examples, we show the results of the program trading based on the "trend reversal" strategy.In the third part, by using the Shanghai and Shenzhen 300 stock-index high-frequency data of futures in a day as test samples, we empirically studied the model validity. The period of the test sample is 1 year(from September 23, 2013 to September 15, 2014)and totally we used 65283 1-minute high frequency data, 12960 5-minutes high frequency data, 4320 15-minutes high frequency data, 2400 30-minutes high frequency data, 1440 60-minutes high frequency data. There are 10201 combinations of the model parameters.