Research On The Statistical Arbitrage In Stock Index Futures Based On The Lasso Method

Under the framework of securities investment, it's a very challenging work to pursuit the stable profit while minimizing the corresponding risk, arbitrage is one of the most efficient methods. In terms of Markowitz's investment portfolio theory, one can get a risk free yield by construct the proper asset portfolio. But with the reality that the securities get higher liquidity, and easier for the mass to come into securities market, this kind of free lunch gets rare. Benefiting from the development of math models and hardware, a new kind of arbitrage strategy-statistical arbitrage emerges out in recent years. This new method doesn't pursuit the risk free yield any more, but to pursuit the stable profit stream along with the existence of tiny risk.The implement of statistical arbitrage makes program trading based on high frequency real time data. The first step is to carry out the short and long position at the same time to get a Risk-neutral portfolio, and this is a important job, Some researches simply construct the Risk-neutral asset from two stocks, this is actually impossible in the stock market by the restriction of short selling stock; some choose all stocks under the stock index futures, this isn't work either because of the great difficulties in execution, the better way is to choose some stocks by stepwise regression. This paper makes a good portfolio using the Lasso regression, which has got great success in gene analysis; the result shows excellent speed and stability in coefficient estimation. The second step in statistical arbitrage is to find out the arbitrage opportunities in misprice data, this paper takes out this step with Co integration method by considering the Heteroscedasticity in misprice series, and execute the arbitrage strategy in in-model period and out-model period respectively to find out the gain, the research shows obvious profit from this method.