| Quantitative trading emerged in 1980s, and it has been an important form of transaction in the current market because of development communication technology and computing technology. In developed markets like American and Europe, volume of transaction through quantitative trading has been above 70% of the total market transaction volume. Correspondingly, in merging country, quantitative trading is rising and will replace the traditional exchange form gradually. By using quantitative method, some investors made excess return, yet lots of investors fell into loss and some of them made big negative influence on market even. Some influential events made quantitative trading come into being a big controversy, like the LTCM's bankrupt in 1998, global financial crisis in 2008 and big slump in American stock at May 6,2010. Although quantitative trading indeed has influenced the market because of quickly trading form, the traders' blind confidence in quantitative trading is the main cause of huge loss and bad effect on market. How to enforce risk management in quantitative trading is the key factor for strategy, and which is becoming an important thing in practice when in turbulent market.At present, the research about quantitative trading main emphasized on the influence of quantitative trading to market and strategy construction. In the field of research on influence of quantitative trading to market, Hendershott and Riordan (2012) and others has proved that quantitative trading has great effect on reducing market friction and transaction cost and improving market liquidity and efficiency. In related to quantitative trading strategy construction, because of advanced computing technology, genetic algorithm and neural network algorithm has been applied to constructing trading strategy. Kapoor et al. (2011) and others have used genetic algorithm to design trading algorithm, Monakhov (2008) and others have introduced neural network model into trading strategy. But there is little literature about risk management in quantitative trading strategy. In practice, the main risk controlling means is to set stop-loss point. Nevertheless, it is difficult to set a suitable stop-loss point, and the method is not objective and lacking timeliness. Under that background, Strub (2012) introduced position management method into quantitative strategy risk management. When compared to stop-loss point way, position management method has advantages in objectivity and time-varying character. Therefore, this paper will construct some strategy with risk controlling based on position management, and then compare the advantages and disadvantages of them.In terms of risk measurement, selecting risk indicator appropriately is the precondition for measuring financial risk effectively. VaR and CVaR are important tail-risk measurement, which are used for practice regularly. CDaR measures the tail-risk based on loss function of drawdown, which is similar to CVaR but has some excellent character than VaR and CVaR. Data is premise for all research, but financial data has been found to be leptokurtosis and fat-tail. And it is important to model the character of leptokurtosis and fat-tail for risk measuring. Extreme value theory (EVT) emphasized on statistical analysis of rare event, which has predominance at fat-tail modeling. The application of EVT in finance has been especially concerned in recent years. But there is still trouble needing to be resolved when using EVT, it is the Scarcity of extreme data. In practice, FHS data simulation is an effective means to dealing with data scarcity.Building trading strategy based on risk controlling needs data fitting, parameter optimization, back-testing, effective data fitting model and appropriate risk indictor. So, this paper will use Shanghai Composite index data from 2006 to 2015, introducing norm analysis and empirical analysis. In norm analysis, this paper will summary the literatures about quantitative trading and risk management, emphasizing on how to introduce risk measuring method into strategy and implement of the newest risk indicator. In empirical analysis, this paper will use GARCH family models fitting Shanghai Composite index return data and FHS simulating data, then use EVT to fit the fat-tail of data. The paper use R packages to compute data, and present research progress and result with tables and charts.The paper constructed three risk controlling strategies, VaR controlled strategy, EVT-VaR controlled strategy, and EVT-CDaR controlled strategy. Then the three strategies are compared by comprehensive dimensions in the paper. This paper has five chapters. First chapter is introduction, present the research background, research significance, research content, and research way and the innovation of the paper. Second chapter is literature survey about research on quantitative trading and risk management. Third chapter is basically theory, which present EVT, risk measurement, FHS.The forth chapter is about strategy construction and empirical analysis. In this chapter, I constructed a trend following strategy as original strategy, and then imported VaR, EVT-VaR, and EVT-CDaR into the strategy. I used Shanghai Composite Index return data in the empirical analysis, and compare the four strategies in some scale. Firstly, I make a test for Shanghai Composite Index return data; found that the data presented fat tail and high peak, so I used three models: ARMA-GARCH, ARMA-EGARCH, ARMA-TGARCH to fitted the data, and selected the best model which was ARMA-TGARCH. Then I used ARMA-TGARCH and FHS to simulate enough data, and it proves the FHS technique is applicable after comparing simulated data to original data. Thirdly, I constructed three strategies based on risk controlling. The paper uses EWMA for building VaR controlled strategy, and by computing an adjusted leverage the strategy realizes risk management. By using EVT and FHS computing VaR, the EVT-VaR strategy is constructed. In the same way, the paper builds EVT-CDaR controlled strategy.According to the research method presented above, there are some results presented. First, the risk leverage adjustment based on current risk the target risk can reflect the risk well. When current risk is low, leverage adjustment is high, meaning investor can increase positon, and vice versa. Second, when target VaR is 3%, strategy based on VaR and EVT-VaR can get more return than original strategy. Third, when target CDaR is set properly, the strategy based on EVT-CDaR also can get more profit than original's, and the higher target CDaR the more profit. Forth, comparing three strategies based on risk controlling, VaR strategy makes highest return, EVT-CDaR strategy makes lowest return but with the lowest back-testing risk. Fifth, when comes to volatility of risk indicator, EVT-CDaR is more stable than VaR.This paper has three innovation points. Firstly, the paper introduced risk management into quantitative strategy constructing, building risk-adjusted leverage with current risk and target risk to adjust position of strategy objectively and time-varyingly. Secondly, I used EVT to fit a distribution for MDD rather than DD, because MDD'tail is fatter and it is more stochastic than DD, which is more appropriate to EVT. Thirdly, it is a beneficial attempt to using FHS to compute CDaR. FHS has advantage in reflecting the characters of original data, which makes it possible to fit an EVT distribution in a short window.Absolutely, there were some deficits in the paper. At the first place, it is not objective at selecting the total sample fitted parameter as rolling window's parameter in the paper. Secondly, the paper selects a trend following strategy as original strategy, which didn't perform well when market fluctuated fiercely. So, further research has two directions. Firstly, it is better to fit data in different windows with different models. Secondly, introducing risk controlling method to a more complex strategy conjugation with market may be better to test the effectiveness and stability of risk measurement in the quantitative trading. |
PDF Full Text Request