基于平稳过程和技术分析的交易策略研究

发布时间：2018-01-01 07:06

本文关键词：基于平稳过程和技术分析的交易策略研究　出处：《华东师范大学》2017年博士论文　论文类型：学位论文

【摘要】：市场有效性假说是传统金融理论的一个基本假设,很多模型的推导都是基于这个基本假设,但近年来越来越多的研究表明,市场有效性在有些情况下是失效的。从行为金融学的角度考虑,市场的无效性可能来源于市场中交易者的心理偏差以及信息不对称性。本文在平稳过程的框架下,尝试挖掘中国市场中非有效的情形,并将其应用在量化投资策略中,以期获得平稳的收益。首先,本文研究了股指期货统计套利策略。协整模型因可以描述了多个资产之间的均衡关系而被广泛地使用在配对套利策略中。协整套利策略可以使用的前提为多个标的资产之间存在稳定的协整关系,但目前为止,该前提在学术界中并没有得到有说服力的理论支撑,而协整关系的失效会导致协整套利策略发生亏损的风险。针对这一问题,本文借鉴协整模型的表达形式,对配对资产价格模型进行了改进,将取对数之后的配对资产价格描述为一个基于滑动平均的滤波项以及剩余的残差项,而且随之说明了,在配对资产价格的对数增量宽平稳时,这个残差项也是宽平稳的,且该结论的成立并不要求标的资产之间存在协整关系。基于上述模型,本文提出了新的套利策略。而后,将新策略的套利收益的来源拆分为具有不确定性的风险项与确定性的收益项,并给出了新模型中配对系数γ的最优参数估计的逻辑与方法。考虑到高频情形下,要求套利策略以最快的速度成交,因此在交易过程中须直接以盘口挂单的对手价价位下单,而若以行情序列的最新价进行下单则无法保证成交。基于以上原因,本文进一步得到了考虑盘口价差的统计套利策略。随后,基于所给出的策略,本文给出了两个成功的实证案例,分别为股指期货跨期套利案例,以及股指期货跨品种套利案例。其中,在股指期货跨期套利案例中,本文指出,在不同频率下,虽然所提出的策略都可以获得正的收益,但使用的数据频率越高,收益风险比越高。另外,为了说明高频套利策略中考虑盘口价差的重要性,本文对比了使用对手价,与使用行情的最新价作为策略成交价时所得到的收益和风险,结果表明,只使用最新价作为策略回测的成交价,会使策略测试得到的收益率虚高,且风险也比实际情况低。然后,本文研究了基于平稳过程理论的单边择时交易策略。在资产价格对数增量服从平稳过程的条件下,基于平稳过程的交易策略具有单位时间对数收益平稳,且具有随着时间增加,单位时间平均收益收敛的优秀性质。这样的性质保证了策略盈利的稳定性和持续性。本文在资产价格对数增量平稳的条件下验证了几个常用技术指标的平稳性,并提出了基于MACD指标的拐点策略,在沪深300股指期货上的回测结果显示,该策略在近连续4年的测试时间内获得了26.86%的年化收益,而Sharpe比率达到2.17。随后提出了基于平稳过程的策略增强方法框架,证明了在增强策略中,为使得增强收益的Sharpe比率最大,须使每个信号的权重配比与标准化的收益预测值成正比,并给出了标准化收益预测值的估计方法。随后,为验证增强策略的有效性,以KD原始策略作为基准,执行所提出的策略增强方法。测试结果表明,原始KD策略的绩效仅为年化收益18.08%,Sharpe比率1.2,而增强后,策略的年化收益变为52.33%,Sharpe比率达到了3.99。为提高策略的收益风险比,本文探讨了构造多策略组合的方法,借鉴Markowitz的均值-方差模型的思想,本文将求解每个子策略的权重转化为一个规划问题的最优解的求解问题,该规划问题以期望收益叠加风险惩罚项为目标函数,以每个权重为非负,且加和为1作为约束项。其后通过实证,验证了权重优化之后的多策略组合,可以大大提高策略的收益风险比。最后,本文研究了四对蜡烛图组合在A股市场中的预测能力。基于Morris的研究,本文首先给出了这四对蜡烛图组合的量化定义。在测试蜡烛图组合短期对于股价的预测效果时,很多学者将每个蜡烛图形态的发生作为一个个相互独立的事件进行统计,导致出现以下问题:一是由于不同股票之间的持仓时间会发生重叠,而导致的股票收益之间的高度正相关;二是该方法忽略了蜡烛图组合在时间轴上分布是否均匀。因此,本文从策略的角度着手分析蜡烛图组合对股价的预测能力,在绝对收益与相对收益两个衡量标准下,分别构造两个股票组合策略,并且证明了当股票价格的对数增量严平稳时,所构造的两个策略的日收益是严平稳的。最后,使用Step-SPA检验对所构造的平稳策略进行有效性测试,并以策略的有效性判断蜡烛图组合的有效性。本文分别在中市值样本和大市值样本下,测试了所构造股票组合策略是否有显著的预测能力。测试结果表明,不同蜡烛图预测能力不同,其中3个蜡烛图组合对不论中市值样本,还是大市值样本都有很强的预测能力。总体来说,蜡烛图组合在中市值样本中的预测能力高于在大市值股票样本中的预测能力。本章所使用的方法也为事件研究提供了一个新的思路与框架。
[Abstract]:Efficient market hypothesis is one of the basic assumptions of traditional financial theory, many models are derived based on the basic assumption, but in recent years, more and more studies show that the effectiveness of market failure in some cases. Considering from the perspective of behavioral finance, market inefficiencies may be derived from the psychological deviation of the market traders and information asymmetry. This paper is based on the stationary process, attempts to explore the non effective Chinese market situation, and its application in quantitative investment strategies, in order to obtain a stable income. Firstly, this paper studies the statistical arbitrage in stock index futures strategy. The cointegration model can describe the equilibrium relationship between a plurality of assets and is widely used in the matching strategy. There are arbitrage cointegration relationship between cointegration between the premise of arbitrage strategies can be used for a plurality of underlying assets, but as of now, The premise in the academic circle has not been a convincing theoretical support, and the failure of the cointegration relationship will lead to cointegration arbitrage risk strategy of loss. To solve this problem, this paper from the expression of cointegration model, paired asset price model was improved, the logarithm of asset prices after the match described as a moving average filter based on the item and the rest of the residuals, but also shows that, in logarithmic incremental asset price stability when paired laterally, the residual is wide stationary, there is a cointegration relationship between the establishment and the conclusion is not a requirement for the underlying assets. Based on the above model, is presented in this paper. The new arbitrage strategy. Then, the source will split the new strategy for arbitrage with uncertainty risk and income uncertainty, and gives the optimal estimation of parameter matching coefficient of new model. The logic and method. Considering the requirement of high frequency case, arbitrage strategies deal with the fastest speed, therefore in the process of transaction shall be directly to the handicap Guadan price match orders, and the latest price if the market sequence orders can guarantee the transaction. Based on the above reasons, this paper further consider the statistical spread bets the arbitrage strategy. Then, the strategy based on, this paper gives two successful examples, respectively, stock index futures and stock index futures arbitrage case, cross species arbitrage case. Among them, the stock index futures arbitrage case, this paper points out that under different frequencies, although the proposed strategy you can get a positive profit, but the data frequency is higher, the higher the ratio of earnings risk. In addition, in order to illustrate the importance of high-frequency arbitrage spreads into bets in senior high school entrance examination, this paper compares the use of rival price, and The latest price of using market price as the strategy of return and risk, the results show that the price of using only the latest price as the strategy back testing, the test strategy obtained unrealistically high yield, and low risk than the actual situation. Then, in this paper a smooth unilateral selection process theory the trading strategy based on the stationary process in asset prices. Obey the logarithmic increment under the condition of stationary process trading strategy based on the unit of time is logarithmic stable income, which has increased with time, excellent properties of unit time average income convergence. The nature of this kind of strategy to ensure the stability of profitability and sustainability. This paper verified the stable the commonly used technical indexes of some asset prices in the logarithmic increment under stationary conditions, and put forward some strategic inflection point based on MACD index, the Shanghai and Shenzhen 300 stock index futures on the back test results show that the strategy Get a 26.86% annualized return on the test time nearly 4 consecutive years, while the Sharpe ratio reached 2.17. and then puts forward the strategy of stationary process enhancement method based on the framework is proved in enhancing strategies, so as to enhance the income maximum ratio of Sharpe to make the weight ratio of each signal with the standard revenue forecast is proportional to the value, and gives a method to estimate value prediction standard yield. Then, in order to verify the validity of the strategy to enhance the original strategy, KD as a benchmark, to enhance the implementation of the proposed method. The test results show that the original performance of KD strategy is only 18.08% annual yield, the ratio of Sharpe is 1.2, and the enhanced strategy, the annual yield is 52.33%, the ratio of Sharpe to 3.99. in order to improve the strategy of income risk ratio, this paper discusses the method of constructing multi strategy combination, using Markowitz's mean variance model thought, this article will Convert the weight of solving each sub strategy for solving an optimal planning problem solution, the planning problem of expected return superposition risk penalty as the objective function, with each weight is nonnegative, and with 1 as the constraint. Then through the empirical study, it is verified that the combined multi strategy weight after optimization, can greatly improve the strategy of income risk ratio. Finally, this paper studies four of candle combination in the A stock market prediction ability. Based on the Morris research, this paper first gives the definition of four quantitative candle combinations. In the test group for short candle price prediction effect, many scholars will occur each candle shape as an independent event statistics, leads to the following problems: one is due to the different stock positions between time will overlap, resulting in the stock returns between the high The degree of positive correlation; two is the method ignores the candle combination on the time axis distribution. Therefore, this paper from the perspective of strategy analysis of candle combination forecasting to stock, in absolute income and relative income two metrics, we construct two stock portfolio strategy, and prove that the logarithmic increment when the stock price strictly stationary, return the two strategy proposed is strictly stationary. Finally, using Step-SPA test to test the validity of the constructed stable strategy, and the strategy to determine the effectiveness of candle combination effectiveness. This paper in the market value of large cap samples and samples the test, that the structure of stock portfolio strategy have significant predictive ability. The test results show that the different candle prediction ability is different, of which 3 Candle combination on whether the market value of the sample, or large cap samples It has strong predictive power. Generally speaking, candle chart combination has a higher predictive power in the middle market value sample than in the large cap stock sample. The methods used in this chapter also provide a new idea and framework for event research.

【学位授予单位】：华东师范大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：F830.9

【参考文献】