vartools 1.0.1

A Python library for different financial calculations and functions, including Value at Risk (VaR), portfolio optimization, and backtesting.

Financial Calculations Library

Overview

This Python library provides functions to calculate the Value at Risk (VaR) and Conditional Value at Risk (cVaR) for financial portfolios, including stock and forex portfolios. These risk measures help in understanding potential losses under given confidence levels. It also allows you to conveniently download price data from Yahoo Finance and perform portfolio optimization using multiple strategies.

Features

• Download stock price data.
• Calculate VaR and cVaR for a stock portfolio.
• Calculate VaR and cVaR for a forex portfolio.
• Supports both long and short positions.
• Outputs results in both percentage and cash value.
• Rebalance a stock portfolio.
• Portfolio optimization for multiple strategies.
• Dynamic portfolio backtest.

Installation

Ensure you have the required dependencies installed:

pip install scipy
pip install numpy 
pip install pandas
pip install yfinance
pip install matplotlib

Then

pip install vartools

Also run

pip install --upgrade vartools

To get the latest version.

Functions

get_data(stocks, start_date, end_date)

A function to download stock data from Yahoo Finance.

Parameters:

---

• stocks : str | list

  The stock tickers to download.

• start_date : str

  The start date for the data in the format YYYY-MM-DD.

• end_date : str

  The end date for the data in the format YYYY-MM-DD.

Returns:

---

data : pd.DataFrame

A DataFrame containing the stock data.

Note: If you prefer to directly download the data from yfinance it is encouraged a format like this:

stocks = ["AAPL", "TSLA", "AMD", "LMT", "JPM"]
data=yf.download(stocks, start="2020-01-01", end="2023-01-01")['Close'][stocks]

Also if you get the data from an excel or csv file create the list stocks or currencieswith the name of the columns in your file for correct functioning. Also make sure to establish yor Datecolumn as index.

---

var_stocks(data, n_stocks, conf, long, stocks)

Calculate the Value at Risk (VaR) and Conditional Value at Risk (CVaR) for a portfolio of stocks.

Parameters:

---

• data : pd.DataFrame

  A DataFrame containing historical stock prices, indexed by date.

• n_stocks : list

  Number of stocks per ticker.

• conf : int | float

  The confidence level for the VaR calculation (e.g., 95 for 95% confidence).

• long : bool

  Indicates the position type:

  • 1 (True) for long positions
  • 0 (False) for short positions

• stocks : list

A list of column names representing the stocks to be included in the portfolio.

Returns:

---

var_stocks_df : pd.DataFrame

A DataFrame containing the VaR and CVaR values both as percentages and in cash terms.

Note: Utilize this function when you have the number of shares of each stock instead of the weights, also n_stocks and stocks must coincide in lenght and order.

---

var_forex(data, positions, conf, long, currencies)

Calculate the Value at Risk (VaR) and Conditional Value at Risk (CVaR) for a portfolio of currencies.

Parameters:

---

• data : pd.DataFrame

  A DataFrame containing historical exchange rates, indexed by date.

• positions : list

  A list of positions for each currency.

• conf : int | float

  The confidence level for the VaR calculation (e.g., 95 for 95% confidence).

• long : bool

  Indicates the position type:

  • 1 (True) for long positions
  • 0 (False) for short positions

• currencies : list

  A list of column names representing the currencies to be included in the portfolio.

Returns:

---

var_df : pd.DataFrame

A DataFrame containing the VaR and CVaR values both as percentages and in cash terms.

Note: n_stocks and stocks must coincide in lenght and order.

---

rebalance_stocks(w_original, target_weights, data, stocks, portfolio_value)

Rebalance a portfolio of stocks to achieve target weights.

Parameters:

---

• w_original : list

  The original weights of the portfolio.

• target_weights : list

  The target weights for the portfolio.

• data : pd.DataFrame

  A DataFrame containing historical stock prices, indexed by date.

• stocks : list

  A list of column names representing the stocks to be included in the portfolio.

• portfolio_value : float

  The total value of the portfolio.

Returns:

---

w_df : pd.DataFrame

A DataFrame containing the original and target weights, as well as the number of shares to buy/sell.

---

var_weights(data, weights, conf)

A function to calculate the Value at Risk (VaR) for a portfolio of stocks.

Parameters:

---

• data : pd.DataFrame

  A DataFrame containing historical stock prices, indexed by date.

• weights : list | np.ndarray

  A list of weights for the portfolio.

• conf : int | float

  The confidence level for the VaR calculation (e.g., 95 for 95% confidence).

Returns:

---

var : float

The VaR value for the portfolio.

Note: It only works for long positions, and the weights must add up to 1.

---

cvar_weights(data, weights, conf)

A function to calculate the Conditional Value at Risk (CVaR) for a portfolio of stocks.

Parameters:

---

• data : pd.DataFrame

  A DataFrame containing historical stock prices, indexed by date.

• weights : list | np.ndarray

  A list of weights for the portfolio.

• conf : int | float

  The confidence level for the CVaR calculation (e.g., 95 for 95% confidence).

Returns:

---

cvar_pct : float

The CVaR value for the portfolio.

Note: It only works for long positions, and the weights must add up to 1.

---

def cvar_contributions(weights, returns, alpha)

A function to calculate the CVaR contributions of each asset in a portfolio.

Parameters:

---

• weights : list | np.ndarray

  A list of weights for the portfolio.

• returns : pd.DataFrame

  A DataFrame containing the returns of the assets in the portfolio.

• alpha : float

  The alpha value for the CVaR calculation (e.g., 0.05 for 95% confidence).

Returns:

---

contributions : list

Note: It is required to write alpha in decimal notation, also this portfolio strategy only works for long positions, so the weights must add up to 1.

---

plot_weights(stocks, weights)

A function to plot the weights of a portfolio.

Parameters:

---

• stocks : list

  A list of stock tickers.

• weights : list | np.ndarray

  A list of weights for the portfolio

Returns:

---

A pie chart showing the portfolio weights.

---

BlackScholes

A class to implement the Black-Scholes model for option pricing and delta hedging.

Methods:

---

• call_delta(S, k, r, sigma, T): Computes the delta of a European call option.
• put_delta(S, k, r, sigma, T): Computes the delta of a European put option.
• delta_hedge(info_call, info_put): Computes the total delta of a portfolio of call and put options.

---

OptimizePortfolioWeights

A class to optimize portfolio asset weights using multiple quantitative portfolio construction techniques based on risk, return, and downside risk measures.

The class supports classical mean–variance optimization as well as downside-risk and tail-risk–based methods commonly used in quantitative finance.

Methods:

---

• opt_min_var() Computes the portfolio weights that minimize total portfolio variance subject to full investment and minimum weight constraints.

• opt_max_sharpe() Computes the portfolio weights that maximize the Sharpe Ratio using expected returns, the covariance matrix, and a risk-free rate.

• opt_min_semivar(rets_benchmark) Computes portfolio weights that minimize target semivariance relative to a benchmark return series, focusing on downside deviations only.

• opt_max_omega(rets_benchmark) Computes portfolio weights that maximize the Omega ratio by balancing upside variability against downside variability relative to a benchmark.

• opt_min_cvar(alpha) Computes portfolio weights that minimize Conditional Value at Risk (CVaR) at a specified confidence level using historical return simulations.

• opt_mcc(alpha) Computes portfolio weights that minimize the maximum individual asset contribution to portfolio CVaR (Minimum CVaR Contribution), promoting tail-risk diversification.

---

DynamicBacktesting

A class to perform dynamic (rolling) backtesting of portfolio optimization strategies over time, using periodic re-optimization of portfolio weights based on historical price data.

This class extends OptimizePortfolioWeights and applies its optimization methods in a realistic backtesting framework, allowing portfolio weights to be recalculated at fixed intervals and portfolio value to evolve day by day.

Methods:

---

• optimize_weights(prices, n_days, periods) Computes optimized portfolio weights for a given rebalancing period using historical price data. The method dynamically updates the inherited optimizer state and returns weights for multiple optimization strategies, including minimum variance, maximum Sharpe ratio, semivariance, Omega, minimum CVaR, and minimum CVaR contribution.

• simulation() Runs a full dynamic backtesting simulation over the specified time horizon. The method:

  • Splits the data into rolling optimization and out-of-sample backtesting windows
  • Rebalances the portfolio at fixed intervals
  • Simulates daily portfolio value evolution for each optimization strategy
  • Returns a time series of portfolio values for all strategies in a single DataFrame

---

Note: See usage examples for better understanding.

Usage Example

import numpy as np
import pandas as pd
import yfinance as yf
import vartools as vt
import matplotlib.pyplot as plt
from scipy.optimize import minimize

get_data

stocks = ["AAPL", "TSLA", "AMD", "LMT", "JPM"]
start_date = "2020-01-01"
end_date = "2023-01-01"

data = vt.get_data(stocks, start_date, end_date)

var_stocks

stocks = ["AAPL", "TSLA", "AMD", "LMT", "JPM"]
start_date = "2020-01-01"
end_date = "2023-01-01"

data = vt.get_data(stocks, start_date, end_date)
n_stocks =[2193, 1211, 3221, 761, 1231]
conf = 95
long = True

var_df = vt.var_stocks(data, n_stocks, conf, long, stocks)
var_df

var_forex

currencies = ['CHFMXN=X', 'MXN=X']
start_date = "2020-01-01"
end_date = "2024-12-02"

data = vt.get_data(currencies, start_date, end_date)
positions = [7100000, 5300000] # How much you have in each currency. Must match the order in currencies.
conf = 99 # Nivel de confianza
long = True

var_forex_df = vt.var_forex(data, positions, conf, long, currencies)
var_forex_df

rebalance_stocks

stocks = ["AAPL", "TSLA", "AMD", "LMT", "JPM"]
start_date = "2020-01-01"
end_date = "2023-01-01"

data = vt.get_data(stocks, start_date, end_date)

rt = data.pct_change().dropna()
stock_value = n_stocks * data.iloc[-1]
portfolio_value = stock_value.sum()
w_original = stock_value / portfolio_value
w_opt = [0.33, 0.15, 0.06, 0.46, 0.00]

rebalance_df = vt.rebalance_stocks(w_original, w_opt, data, stocks, portfolio_value)
rebalance_df

var_weights

stocks = ["AAPL", "TSLA", "AMD", "LMT", "JPM"]
start_date = "2020-01-01"
end_date = "2023-01-01"

data = vt.get_data(stocks, start_date, end_date)

weights = [0.2457, 0.1301, 0.1820, 0.3064, 0.1358]
conf = 95
var_pct = vt.var_weights(data, weights, conf)
var_pct

cvar_weights

stocks = ["AAPL", "TSLA", "AMD", "LMT", "JPM"]
start_date = "2020-01-01"
end_date = "2023-01-01"

data = vt.get_data(stocks, start_date, end_date)

weights = [0.2457, 0.1301, 0.1820, 0.3064, 0.1358]
conf = 95
cvar_pct = vt.cvar_weights(data, weights, conf)
cvar_pct

cvar_contributions

# bonds, commodities, equities and real estate
stocks = ['VBTLX', 'GSG', 'VTI', 'VNQ']
start_date = '2019-01-01'
end_date = '2024-01-01'

data = vt.get_data(stocks, start_date, end_date)
returns = data.pct_change().dropna()
alpha = 0.05

mcc_weights = vt.mcc_portfolio(returns, alpha)

cvar_contributions = vt.cvar_contributions(mcc_weights, returns, alpha)
cvar_contributions

plot_weights

stocks=['WMT','AAPL','GOOGL','PG','XOM','KO','CMG','F']
start_date='2020-01-01'
end_date='2024-11-24'

data = vt.get_data(stocks, start_date, end_date)
returns = data.pct_change().dropna()
rf = 0.04413

min_var = vt.OptimizePortfolioWeights(returns, rf).opt_min_var()

vt.plot_weights(stocks, min_var)

call_delta and put_delta

S_call = 20.3
K_call = 20.43
r_call = 0.0425
sigma_call = 0.102
T_call = 1/12

S_put = 20.3
K_put = 20.2
r_put = 0.0425
sigma_put = 0.156
T_put = 1/12

delta_call = vt.BlackScholes().call_delta(S_call, K_call, r_call, sigma_call, T_call)
delta_put = vt.BlackScholes().put_delta(S_put, K_put, r_put, sigma_put, T_put)

delta_call, delta_put

# Write in order S, K, r, sigma, T

call = [20.3, 20.43, 0.0425, 0.102, 1/12]
put = [20.3, 20.2, 0.0425, 0.156, 1/12]

delta_call = vt.BlackScholes().call_delta(*call)
delta_put = vt.BlackScholes().put_delta(*put)

delta_call, delta_put

delta_hedge

# Write in order S, K, r, sigma, T, N (money invested in each option)

info_call = [[20.3, 20.43, 0.0425, 0.102, 1/12, 23],
            [20.3, 20.52, 0.0425, 0.111, 1/12, 25],
            [20.3, 20.43, 0.0421, 0.297, 6/12, 17],
            [20.3, 20.52, 0.0421, 0.289, 6/12, 32]]

info_put = [[20.3, 20.2, 0.0425, 0.156, 1/12, 12],
            [20.3, 20, 0.0425, 0.153, 1/12, 16],
            [20.3, 20.2, 0.0421, 0.348, 6/12, 11],
            [20.3, 20, 0.0421, 0.378, 6/12, 17]]

# If N is in millions of dollar, then
hedge = vt.BlackScholes().delta_hedge(info_call, info_put)
print(f'Buy {hedge} millions of dollars of the underlying asset')

OptimizePortfolioWeights

stocks=['WMT','AAPL','GOOGL','PG','XOM','KO','CMG','F']
start_date='2020-01-01'
end_date='2024-11-24'

data = vt.get_data(stocks, start_date, end_date)
returns = data.pct_change().dropna()
rf = 0.04413

opt_sharpe_weights = vt.OptimizePortfolioWeights(returns, rf).opt_max_sharpe()
opt_sharpe_weights

tickers = ['NVDA','AMZN','AVGO','PG','V','RL','GLD']

benchmark = 'SPY'
start_date = '2025-01-01'
end_date = '2025-12-31'

price = vt.get_data(tickers, start_date, end_date)
benchmark_data = vt.get_data([benchmark], start_date, end_date)

rt_benchmark = benchmark_data.pct_change().dropna()
rt = price.pct_change().dropna()

min_semivar_weights = vt.OptimizePortfolioWeights(rt, 0.0).opt_min_semivar(rt_benchmark)
min_semivar_weights

DynamicBacktesting

stocks=['WMT','AAPL','GOOGL','PG','XOM','KO','CMG','F']
start_date='2020-01-01'
end_date='2024-11-24'

data = vt.get_data(stocks, start_date, end_date)
returns = data.pct_change().dropna()
rf = 0.0035
pv = 1_000_000.0
months = 2

history = vt.DynamicBacktesting(price, benchmark_data, capital=1_000_000, rf=rf, months=months).simulation()
history

License

This project is licensed under the GPL-3.0 license.

Author

Luis Fernando Márquez Bañuelos