longtailfinancial / Options-Project

Research on derivatives (American options) using machine learning algorithms trained on historical data.

Options Project

Description

We aim to design and evaluate the performance of various option pricing models. Alpha Vantage and Questrade API retrieve historical and current stock prices, respectively. Latter also scrapes for real-time option spreads. Discount Option Data is used to obtain historical option spreads. FRED is responsible for treasury yields.

Users are responsible for obtaining their own API keys.

Overview

• scrape_and_preprocess.py:

  • Retrieve historical closing prices using Alpha Vantage
  • Adjust for splits
  • Estimates dividend contribution to price

• scrape_options.py:

  • Filter options data for specified ticker and append dividend data

• scrape_treasury_yields.py:

  • Retrieve treasury yields from Federal Reserve Economic Data (FRED)

• greeks.ipynb (here if GitHub is unable to load)

  • Derive Delta, Gamma and Theta using option spread data
  • Interpolate above greeks for strike prices in option spread

• Exploratory Data Analysis

  • EDA_1.ipynb (here)
    • Wrangle, filter, process, and visualize adjusted options data
    • Visualization of open interest
  • EDA_2.ipynb (here)
    • Engineer novel features
    • De-trend target with ARIMA. Ensure stationary using Augmented Dickey-Fuller test
    • Baseline linear regression model
    • Principal component analysis (PCA) on features

Data

• EDA

  • Format: {ticker}_(calls/puts)_EDA1.csv
  • Open interest of call/put options. Filters out newly created and expired options

• greeks

  • Format: {ticker}_(calls/puts)_(delta/gamma/theta).csv
  • greek values for calls/puts derived from option spread. See greeks for details

• adjusted_daily_closing

  • Format: {ticker}.csv
  • End-of-day prices from Alpha Vantage. "close" column prices have been forward/reverse split adjusted. "adjustment factor" column is the factor by which the raw unadjusted price was divided by

• adjusted_options

  • Format: {ticker}.csv
  • Adjusted end-of-day option spreads from Discount Option Data

• dividends

  • Format:
    • {ticker}.csv: The start/end dates of each dividend period, and amount paid. Note that "div_start" is the ex-dividend date for the previous dividend
    • {ticker}_ts.csv: Priced-in dividends as a time series (daily). Function "dividend_pricing" uses a linear pricing model
  • For each ticker, there are two files. All dividends have been split adjusted, and can be subtracted from data in "adjusted_daily_closing" to obtain true equity price

• questrade

  • Format: {ticker}/(date).csv
  • Unadjusted option data scraped from Questrade API in real-time. Thus, recorded prices may not represent end-of-day
  • Each ticker has its own folder containing options spreads expiring on different days
  • TODO: record date of scrape, wrangle into same format as training data

• treasury_yields

  • Format: {time_to_maturity}.csv
  • US treasury yield (return ratio per day). Unit of day rather than year to match greeks (e.g. Theta)
    • daily rate by linear interpolation
    • daily rate by exponential interpolation (continuous compounding).