Tutorials

HurstHalfLifeCointegration()

Analyzes stock pairs based on cointegration, Hurst exponent, half-life, and mean crossings.

The HurstHalfLifeCointegration class is designed to analyze stock pairs by fetching their historical price data and applying a series of statistical tests. The class uses the Engle-Granger cointegration test, calculates the Hurst exponent, estimates the half-life of the spread between the pairs, and counts the number of mean crossings, based on configurable thresholds.

Class: HurstHalfLifeCointegration()

A tool for identifying mean-reverting pairs of stocks using statistical methods.

Generates a list of eligible pairs based on:

  • Cointegration test

  • Hurst exponent (for mean reversion check)

  • Half-life of the spread

  • Mean crossings

Supports customizable thresholds for the selection of eligible stock pairs based on user-defined parameters.

Constructor: __init__()

Initializes the HurstHalfLifeCointegration object and configures thresholds for analysis.

Arguments:

  • tickers (list): List of stock tickers to analyze.

  • start (str): Start date for fetching historical stock data in the format 'YYYY-MM-DD'.

  • end (str): End date for fetching historical stock data in the format 'YYYY-MM-DD'.

  • p_value_threshold (float, optional): Maximum acceptable p-value for the cointegration test. Default is 0.05.

  • hurst_threshold (float, optional): Maximum acceptable Hurst exponent value (less than 0.5 for mean-reversion). Default is 0.5.

  • min_half_life (int, optional): Minimum half-life of the spread in days. Default is 1.

  • max_half_life (int, optional): Maximum half-life of the spread in days. Default is 365.

  • min_crossings (int, optional): Minimum number of mean crossings per year. Default is 12.

Attributes Initialized:

  • tickers: The list of stock tickers to analyze.

  • start: The start date for downloading stock data.

  • end: The end date for downloading stock data.

  • p_value_threshold: The maximum p-value for the cointegration test.

  • hurst_threshold: The maximum Hurst exponent value to filter mean-reverting pairs.

  • min_half_life: The minimum half-life of the spread.

  • max_half_life: The maximum half-life of the spread.

  • min_crossings: The minimum number of mean crossings required.

  • data: Holds the fetched stock data.

Method: download_data()

Downloads historical stock data for the specified tickers using the Yahoo Finance API (yfinance).

Arguments:

None

Returns:

None (Stock data is stored in the self.data attribute)

Method: calculate_cointegration()

Performs the Engle-Granger cointegration test on a pair of stocks.

Arguments:

  • pair (tuple): A tuple containing two stock tickers.

Returns:

  • A tuple of the two tickers and their p-value if the p-value is below the p_value_threshold.

  • None if the p-value is above the threshold or an error occurs.

Method: calculate_hurst()

Calculates the Hurst exponent to determine if a stock pair exhibits mean-reverting behavior.

Arguments:

  • series (pd.Series): A time series of the spread between two stocks.

Returns:

  • The Hurst exponent (float) if the series is valid and meets the criteria.

  • None if the calculation fails due to invalid data or an error.

Method: calculate_half_life()

Calculates the half-life of mean reversion for a stock pair spread using an Ordinary Least Squares (OLS) regression.

Arguments:

  • spread (pd.Series): The spread between two stock prices.

Returns:

  • The half-life of the spread (float).

Method: calculate_mean_crossings()

Counts how many times the spread crosses its mean, which helps identify whether the pair exhibits frequent mean reversion.

Arguments:

  • spread (pd.Series): The spread between two stock prices.

Returns:

  • The number of mean crossings (int).

Method: process_pair()

Processes each pair of stocks by:

  1. Checking cointegration.

  2. Calculating the Hurst exponent.

  3. Estimating the half-life of the spread.

  4. Counting the mean crossings.

Arguments:

  • pair (tuple): A tuple of two stock tickers.

Returns:

  • A tuple of the two tickers and their statistical properties (p-value, Hurst exponent, half-life, mean crossings) if they meet all the thresholds.

  • None if the pair fails any of the tests or thresholds.

Method: find_eligible_pairs()

Applies the process_pair method to all possible combinations of the stock tickers to find eligible pairs using parallel processing.

Arguments:

None

Returns:

  • A list of eligible stock pairs that meet all the statistical criteria.

Method: run_analysis()

Downloads the stock data and processes all pairs to find eligible pairs.

Arguments:

None

Returns:

None

Example Usage

tickers_csv_path = 'tickers.csv'
tickers = load_tickers_from_csv(tickers_csv_path)

# Initialize the analyzer with customizable parameters
analyzer = HurstHalfLifeCointegration(
    tickers=tickers,
    start='2022-01-01',
    end='2024-01-01',
    p_value_threshold=0.05,
    hurst_threshold=0.4,
    min_half_life=5,
    max_half_life=250,
    min_crossings=10
)

# Run the analysis
analyzer.run_analysis()

Customization Options

  • Thresholds for Statistical Tests: Users can set custom thresholds for cointegration p-value, Hurst exponent, half-life of mean reversion, and mean crossings to fine-tune the analysis.

  • Stock Tickers: The class accepts any combination of tickers supported by Yahoo Finance.

  • Date Range: Allows specifying a date range (start and end) for historical data analysis.

PreviousMeanAnalyzer()NextGreek Exposure

Last updated