Portfolio Rebalancing Optimization

A stock price simulation that optimizes the rolling net sharpe ratio of a portfolio. The model supports several heuristic optimization algorithms.

1. Features

Sandbox

main_rebalancing.py is a little sandbox that let's you play around with the parameters. The underlining data is from the german stock market (germanstocks.csv). The data is being bootstrapped to ensure a different outcome for every simulation. The objective function is the effective (net) sharpe ratio after transaction costs.

Experimental Setup

experiment_rebalancing.py is an experimental design to try several hyperparameters. Multi processing is included. The script writes a .csv and evaluates several performance measures.

Model Parameters

Currently the implementation support several parameters.

• strategy is the algorithm that optimizes the objective function which is the effective (or net) sharpe ratio. Possible values are:
  1. Differential Evolution Alorithm on the portfolio weights x
  2. Differential Evolution Algorithm on the delta of portfolio weights dx
  3. A simple neighborhood search
  4. Sequential Least Squares Programming from scipy
  5. Equally weightd portfolio strategy
• perfectForesight determines if the covariance matrix is estimated based on the true data generating process (True) or on past observations (False).
• rS is the interest (daily) rate of the risk-free asset.
• w0 is the starting budget.
• T_prior is the rolling time horizon used for the estimation of the covariance matrix.
• T_invst is the investment horizon.
• BL is the block length for the bootstrap.
• cvar is the amount of variable transaction costs.

Examples

Example of a three year portfolio simulation based on Sequential Least Squares Programming optimization. The plot shows: ( a ) Stock market prices ( b ) Portfolio weights over time ( c ) portfolio balance the dynamic portfolio (net/gross) and a portfolio without rebalancing. ( d ) Rolling sharpe ratio w.r.t. to marturity

Example of an "equally weighted portfolio"-strategy:

2. Dependencies

• numpy
• matplotlib
• pandas
• scipy

To install requirements, cd to the directory of the repository and run pip install -r requirements.txt. A virtual environment is recommended.

3. TO DO

• other evolutionary optimization algorithms, i.e. Particle Swarm Optimization
• other objective functions

---

4. Contact

I am very thankful for any kind of feedback. Also, if you have questions, please contact gregor.lenhard@unibas.ch .