Sdoof / StatArbPairsTrading

This repository includes an introduction to statistical arbitrage pairs trading. Specifically, I discuss some of the research methods required in order to find a successful pair as well as the code implementation for a backtest.

StatArbPairsTrading

This repository includes an introduction to statistical arbitrage pairs trading. My jupyter notebook, StatArbBlog.ipynb, is divided into two central categories: research and backtesting.

Research

In the research aspect of my jupyter notebook, I discuss a few different statistical approaches to analyze cointegration and mean-revresion. Additionally, I discuss a rather simple way to finding a pair with cointegration tendencies, i.e., selected the most liquid competing companies within a specific sectior. In my example, I used the car manufacturing industry as my sector of choice.

Backtesting

In the backtesting aspect of my jupyter notebook, I discuss how to formally create a backtest as well as calculate realized pnl and sharpe ratio. By no means do I consider this way to be the best way of creating a backtest, I've just had great success with this method and it has allowed me to calculate performance metrics rather easily. Additionally, I did not discuss calculating unrealized pnl, drawdowns, or rolling sharpe ratio. Again, this notebook is to get the reader accustomed to developing simple statistical arbitrage pairs trading strategies.

Notice

I'd also like to mention that there are MANY things I that have excluded from my baktest. Some of which include: calculating market impact, analyzing daily volume for both stocks in my pair to find an ideal and accessible number of shares to trade with, using minute market data instead of daily data, taking corporate actions into account (e.g. dividends), etc.