Backtesting trading strategies

Last Update February 25, 2021 ####

Matteo Bottacini, matteo.bottacini@usi.ch ####

Project description

This project is to backtest different trading strategies applying different approaches from the Modern Portfolio Tehory (MPT) in Python 3.

The strategies backtested are:

• The Optimal Markowitz Portfolio;
• The Global Minimum Variance Portfolio;
• The Risk-Parity Portfolio;
• The Equally Weighted Portfolio

The ETFs considered are:

• EEM: iShares MSCI Emerging Markets ETF
• EMLC: VanEck Vectors J.P. Morgan EM Local Currency Bond ETF
• IAU: iShares Gold Trust
• IEF: iShares 7-10 Year Treasury Bond ETF
• IWM: iShares Russell 2000 ETF
• SPY: SPDR S&P 500 ETF Trust
• TIP: iShares TIPS Bond ETF
• TLT: iShares 20+ Year Treasury Bond ETF
• VGK: Vanguard FTSE Europe Index Fund ETF Shares

The scripts do the following:

• Download and analyse financial data;
• Find the optimal Markowtiz portfolio (mean-variance);
• Find the Global Minimum Variance (GMV) portfolio (minimum variance);
• Find the risk-parity portfolio (risk parity);
• Find the equally weighted portfolio (equally weighted);
• Backtest a trading strategy with monthly rebalance with out-of-the-sample results;
• Compare the results.

Folder structure: ~~~~ trading-strategy-backtest/ deliverable/ run_backtest.py src/ equallyweightedportfolio.py meanvarianceportfolio.py minimumvarianceportfolio.py riskparityportoflio.py README.md ~~~~

Trading strategy configuration

Equally weighted portfolio ###

def equallyweightedportfolio(ret):
    init_weights = [1 / len(ret.columns)] * len(ret.columns)
    optweights = initweights
return opt_weights

Risk parity portfolio ###

def riskparityportfolio(ret):
    init_guess = 1 / ret.std()
    optweights = list(initguess / init_guess.sum())
return opt_weights

Minimum variance portfolio ###

import numpy as np
from scipy.optimize import minimize
def minimumvarianceportfolio(ret):
# define objective function to minimize: variance     def getportfoliovariance(weights):         weights = np.array(weights)  # check         cov_mat = ret.cov()         portvariance = np.dot(weights.T, np.dot(covmat, weights))         return port_variance
# equality constraint: sum of the weights = 1     def weight_cons(weights):         return np.sum(weights) - 1
# model set-up     # - long only portfolio     # - initial guess     # - constraints     bounds_lim = ((0, 1),) * len(ret.columns)     init_weights = [1 / len(ret.columns)] * len(ret.columns)     constraint = {'type': 'eq', 'fun': weight_cons}
# find optimal portfolio     optport = minimize(fun=getportfolio_variance,                         x0=init_weights,                         bounds=bounds_lim,                         constraints=constraint,                         method='SLSQP')
# find optimal weights     optweights = list(optport['x'])
return opt_weights

Mean-variance portfolio ###

# import modules
import numpy as np
from scipy.optimize import minimize
def meanvarianceportfolio(ret):
# define objective function to minimize: sharpe ratio     def getportfoliosr(weights):
weights = np.array(weights)  # check
# expected returns         port_ret = np.dot(ret, weights)         meanret = portret.mean()
# volatility         cov_mat = ret.cov()         portstd = np.sqrt(np.dot(weights.T, np.dot(covmat, weights)))
# sharpe ratio         portsr = meanret / port_std         return port_sr
def objective_fun(weights):         negsr = getportfolio_sr(weights) * (-1)         return neg_sr
# equality constraint: sum of the weights = 1     def weight_cons(weights):         return np.sum(weights) - 1
# model set-up     # - long only portfolio     # - initial guess     # - constraints     bounds_lim = ((0, 1),) * len(ret.columns)     init_weights = [1 / len(ret.columns)] * len(ret.columns)     constraint = {'type': 'eq', 'fun': weight_cons}
# find optimal portfolio     optport = minimize(fun=objectivefun,                         x0=init_weights,                         bounds=bounds_lim,                         constraints=constraint,                         method='SLSQP')
# find optimal weights     optweights = list(optport['x'])
return opt_weights

Backtest configuration

• Assets;
• Length of the training set

# feel free to change the following parameters:
tickers = []
monthstrainingset = 12 * 5

Results

Supported versions