As part of an ongoing effort to provide finance-related Python libraries, I will start with portfolio optimization. This page documents the library’s Hello-World version.
Update: the library now also runs in the browser. The Portfolio Optimizer loads this tutorial’s five factors, bounds, and views so you can change them and watch the frontier move.
Installation
If you have Python 3.6+ installed, run the following in your terminal:
pip install git+https://github.com/WizardKingZ/portfolio_optimization.git
To uninstall it, run:
pip uninstall portfolio_optimization
Getting Started
Unconstrained Portfolio
## import MarkowitzPortfolio
from portfolio_optimization import MarkowitzPortfolio
import numpy as np
## Load the Fama-French Five Factor Dataset to calculate annualized return and covariance
rts = np.array([0.11, 0.07, 0.09, 0.08, 0.08])
cov = np.array([[ 0.0242, -0.0015, -0.0025, -0.0019, -0.0033],
[-0.0015, 0.0067, 0.0004, -0.0013 , 0.0001 ],
[-0.0025, 0.0004, 0.0063, -0.0002, 0.0025],
[-0.0019, -0.0013 , -0.0002, 0.0033, 0.0001],
[-0.0033, 0.0001 , 0.0025, 0.0001, 0.0033]])
ffFactorNames = ['Market', 'SMB', 'HML', 'RMW', 'CMA']
## Initialize the MarkowitzPortfolio with the expected return, covariance and asset names
## both cov and rts are numpy arrays. rts is a row vector
## ffFactorNames should be a list
port = MarkowitzPortfolio(rts, cov, ffFactorNames, riskFreeRate=0.046)
## set the target return as 0.1
## you can set risk aversion as well
## we will cover constraints optimization later
configuration = {'constraints': None,
'riskAversion': None,
'targetReturn': 0.1}
## obtain weights allocated on each asset
weights = port.get_allocations(configuration=configuration)
## plot the efficient frontier and capital market line
port.display_efficient_frontier(assetsAnnotation=True,
specialPortfolioAnnotation=True,
addTangencyLine=True,
upper_bound=0.14)

Constrained Portfolio
Continuing with the Fama-French five-factor (FF-5) data, suppose you want a portfolio with the following constraints:
\[\begin{array}{c|c|c} \textbf{Asset} & \textbf{Lower Bound %} & \textbf{Upper Bound %}\\\ \text{MKT} & 0 & 100 \\\ \text{HML} & 50 & 100\\\ \text{SMB} & 0 & 100 \\\ \text{RMW} & 0 & 100 \\\ \text{CMA} & -50 & 100 \\\ \end{array}\]Set up these constraints as follows:
constraints = {'Market': [0, 1],
'HML': [0.5, 1],
'SMB': [0, 1],
'RMW': [0, 1],
'CMA': [-.5, 1]}
configuration['constraints'] = constraints
## obtain weights allocated on each asset
weights = port.get_allocations(configuration=configuration)
Black-Litterman Portfolio Allocation
Investors often want to incorporate their own views when solving asset allocation problems. For example:
\[\begin{array}{c|c|c} \textbf{View} & \textbf{Confidence Level %} & \textbf{Plus and Minus %}\\\ \text{Market outperforms by 1%} & 95 & 5 \\\ \text{SMB beats RMW by 1%} & 90 & 5\\\ \text{A portfolio of 20% of SMB and 80% of HML beats RMW by 1%} & 99 & 1\\\ \end{array}\]Use BlackLittermanPortfolio to incorporate these views:
from portfolio_optimization import BlackLittermanPortfolio
## assetNames are ['Market', 'HML', 'SMB', 'RMW', 'CMA']
port = BlackLittermanPortfolio(modeledReturn, modeledCovariance, assetNames, riskFreeRate=0.046)
views = {'Market': {'type': 'absolute', 'scale': 0.01, 'confidence': 95, 'plusminus%': 5},
'SMB|RMW' : {'type': 'relative', 'scale': 0.01, 'confidence': 90, 'plusminus%': 5, 'weights': [1., -1.]},
'SMB|HML|RMW' : {'type': 'relative', 'scale': 0.01, 'confidence': 99, 'plusminus%': 1, 'weights': [.2, .8, -1.]},
}
## second parameter is the R^2 used in estimating your modeled expected returns
port.update_views(views, 0.1)
## Let's use the same parameters
configuration = {'constraints': None,
'riskAversion': None,
'targetReturn': 0.1}
## you can obtain the BL weights
weights = port.get_allocations(configuration=configuration)
## you can reset the portfolio to become a regular Markowitz portfolio
port.reset()
The second argument of update_view is the \(R^2\) of your modeled returns; I will add factor modeling tools in a later iteration. As a starting point, consider the most widely used factor model, the Capital Asset Pricing Model (CAPM). CAPM-estimated returns typically have an \(R^2\) below 15%, so set this parameter below 15% (the example above uses 10%).
I will add more functionality over time. The next iteration will include:
- Hansen-Jagannathan bound test
- Factor models for estimating stochastic discount factors
- Style analysis