An Empirical Study on the Effectiveness of Technical Analysis
|School||Southwestern University of Finance and Economics|
|Keywords||Technical Pattern Analysis Test Based on the Simulation Effectiveness Kernel Regression|
Technical analysis, also known as "charting", has been a part of financial practice for many decades, but there have been arguments with its principles and methods while investors have been using it. The focus of the dispute is that technical analysis has a highly subjective nature in the operation.In this paper, we hope to bridge the gulf between technical analysis and quantitative finance.We propose a systematic and automatic approach to technical pattern recognition using nonparametric kerne regression, and we apply this method to a large number of thirty stocks in Shanghai Stock Exchange and Shanghai composite index to evaluate the effectiveness of technical analysis.First, constructing kernel estimator of a given time series of daily closing prices of each stock.And then finding the local extreme on the price.Second, define ten kinds of technical patterns using a sequence of consecutive extreme,so that we can now construct an algorithm for automating the detection of technical patterns.We use the algorithm in the sample of thirty stocks and Shanghai composite index, analyze for occurrences of each technical pattern and statistic feature of conditional distribution.Third,we compare the unconditional empirical distribution of daily stock returns to the conditional distribution by using the test based on the simulation.The empirical results shows:head-and-shoulders and inverse head-and-shoulders are the most frequently occurring.The second is broadening tops and bottoms. Double tops and bottoms are the least number of occurrences.For returns of stock, pattern of triangle bottoms and head-and-shoulders have the highest return.Triangle bottoms have higher conditional returns in twenty stocks than other patterns.The trade indicators come from triangle bottoms could bring the investor excess return.The pattern of triangle bottoms is effective.In the other hand, head-and-shoulders appear many times,but this pattern only have higher conditional returns in nine stocks. Therefore, we believe triangle bottoms are more effective than head-and-shoulders.We find broadening tops and bottoms, rectangle tops and bottoms, and double bottoms have higher conditional returns in a few stocks.This fact show that when we invest in a particular stock, the five pattern we talk above are effective.There are three patterns have no effectiveness in all of stocks, they are inverse head-and-shoulders, triangle tops and double tops. Furthermore, we find that all of patterns are noneffective in four stocks.Above all, triangle tops and head-and-shoulders are effective in shanghai stock market.but we also notice, the technical patterns are not effective in all stocks. Therefore, we conclude that using this approach alone is not enough,we should provide other kinds of method of investment analysis. Using a variety of methods is best.There is one innovation point of this paper.We compare the unconditional empirical distribution of daily stock returns to the conditional distribution by using the test based on the simulation. The principle of this tese is, comparing investors of technical analysis with the random trader, we judge whose return of stock is higher.The procedure is, first,we make m random trader, calculate the mean of returns of them.these date are a sample;second,we compare the sample to the conditional returns.If the conditional returns are higher than95%quantile of this sample,we deside technical patterns are effective.There are many advantages in the tese.For example,principle of test is simple and clear. We could use this test between Kolmogorov-Smirnov test.The conclution will be more convictive.This paper also has some shortcomings.The first shortcoming is that,we do not consider long-term patterns.Because we define patterns in a fixed windows of38days,of couse,we can only find patterns that are completed within38trading days.This is a restriction of this paper.The second shortcoming is we do not consider the effect of trading volume.In fact, trading volume is a important factor in technical analysis.The third shortcoming is we do not consider trade cost. In the futher study, we will take these factors into the model, to reseach more general conclutions.