This repository contains code for a mean-reversion trading algorithm based on the Ornstein-Uhlenbeck (OU) process. The scripts in this repository are meant to illustrate the implementation of such an algorithm and should not be used to inform real investment decisions.
This code requires Python3 and certain external libraries ("dependencies") to run. First, make sure you have Python3 installed in your system. Then, install the dependencies by running
pip install -r requirements.txt
There are two executale scripts in this repository, mean_reversion.py and market_performance.py.
mean_reversion.py reads a list of companies, checks them for mean-reverting behavior, fits an OU process to their price series, and prints a list of companies whose stock can be potentially bought or shorted. This script outputs recommended portfolio allocations for each company too. This program takes four command-line arguments: the file name containing the company list, a start date and end date for the look back period in the format yyyy-mm-dd, and a threshold for the Augmented Dickey-Fuller test, which filters out non-mean reverting companies from the original list. For example, consider the command
python3 mean_reversion.py companies.in 2023-01-01 2024-12-31 0.05
which reads a list of companies from companies.in, selects the daily close prices of these companies from January 1, 2023 to December 31, 2024, and uses p < 0.05 to identify companies whose prices don't exhibit mean-reverting behavior.
market_performance.py reads a list of companies and their portfolio allocations and plots the performance of the portfolio against the S&P 500 and NASDAQ over a number of years. It takes four command-line arguments: the file name containing a list of companies and the percentage of each one within a portfolio, a start and end date for the lookback period in the format yyyy-mm-dd, and an amount of starting capital. For example, running
python3 market_performance.py portfolio.in 2020-06-01 2024-12-01 10000
creates a plot of the performance of the portfolio in portfolion.in from June 1, 2020 to December 1, 2024 against the NASDAQ and S&P 500 with a starting capital of $10,000. This program does not support intra-year lookback periods, meaning the start and end dates must have different years.
For mean_reversion.py and market_performance.py to run properly, they expect inputs in specific formats.
mean_reversion.py expects a file where each line contians a stock symbol and the name of the corresponding company separated by whitespace (companies.in illustrates this format).
market_performance.py expects a file where each line contains a stock symbol, the name of the company, and the company's percentage in the portfolio (take a look at portfolio.in for an example).