Statistical Arbitrage!

With my experience in statistical analysis and after further gaining information on equity markets I was curious to explore the possibility of generating alpha by combining the two skills. I got an opportunity where I worked on a similar idea and then I extended it on the data I had downloaded from yahoo.
In this strategy I try to find a statistical equivalent of pairs trading and try to extract inefficiencies in price movements.

Abstract

The objective of this analysis is to obtain an alpha that is based on statistically exploring ineffi ciencies in stock prices. The strategy involves decomposing stock prices in each industry into principal components that explain the most variance and then regress the stock prices on those components to obtain the stock's  dependence on them. Next the components are forecasted using GARCH model and hence the forecasted evolution of the stocks is also obtained based on the regression results. Based on these forecasts I will create a long-short neutral arbitrage strategy with the aim of achieving high risk adjusted returns.

Statistical Arbitrage on US Equity
https://github.com/kunalrajani/statistical-arbitrage

Disclaimer: Please read
Please note that this is my independent work where I have used data from yahoo finance to explore statistical concepts from a course I took in statistics. It may inadvertently have an overlap with a work that somebody else has already done and I have no intentions of replicating it. I would be glad to know of any such clash and post a clarification on this post.
I am open to having my work being redistributed or used but only after due credit and a reference has been made. Feel free to contact me to avoid any misunderstandings or if you need more details from this paper. 

*The analysis is still underway and the conclusions are under review

Understanding Factor models and Cone programming

This was a very interesting project we undertook. Not because its result was any enormous value addition but because it helped in grasping concepts and procedures critical to function in practice. It helped in going through the rigorous process of data cleaning and implementing factor models and checking their significance to predict stock returns. Further on we explored optimization using cone programming, which is a special case of interior point methods, to maximize the sharpe ratio and obtain a market beating portfolio.

Here is the Abstract:

The goal of our project is to utilize factor models to explain returns and optimize the Sharpe ratio to create a portfolio that outperforms the S&P 500. After re fining our data we have a universe of 335 stock in which we can invest. We re-balance our portfolio quarterly and incorporate factor models and Sharpe ratio optimization through cone programming to form the portfolio. The rest of the paper is organized as follows: Section 1 is a short introduction of our paper, Section 2 gives a brief idea of the data available and what kind of choices we made to reach the final universe of stocks, Section 3 gives an idea of the general methodology used in the paper, Section 4 describes the results that we have reached, Section 5 presents the significance test we performed, Section 6 presents the results of different sensitivity analysis and Section 7 summarizes the project and gives suggestions for further research.

Investment Allocation using Factor models and Cone programming optimization