Pandas & NumPy Workshop — Signal and Backtesting using Vectors

Rutgers Quant Finance Club

A hands-on workshop where you build a Mean-Reversion trading strategy from scratch using Pandas for signal generation and pure NumPy for high-performance vectorized backtesting.

---

What You'll Learn

Concept	Tools
Fetching & cleaning market data	yfinance, pandas
Rolling statistics & Z-Score signals	pandas.rolling()
Trade signal generation (Long / Short / Flat)	numpy.where()
Avoiding look-ahead bias with .shift()	pandas
Vectorized backtesting with log returns	numpy
Sharpe ratio calculation	numpy
Equity curve visualization	matplotlib

Repository Structure

pandas_numpy_workshop/
├── workshop.ipynb            # Student worksheet (fill-in-the-blank exercises)
├── workshop_solution.ipynb   # Completed solution notebook
└── README.md

• workshop.ipynb — The main worksheet. Code cells contain scaffolded exercises with hints and comments guiding you to complete each step.
• workshop_solution.ipynb — The fully completed version for reference after attempting the workshop.

Getting Started

Prerequisites

• Python 3.10+
• Jupyter Notebook or JupyterLab

Installation

# Clone the repo
git clone https://github.com/Rutgers-Quant-Finance-Club/pandas_numpy_workshop.git
cd pandas_numpy_workshop

# Install dependencies
pip install pandas numpy matplotlib yfinance jupyter

Running the Workshop

jupyter notebook workshop.ipynb

Or Open in GitHub Codespaces

No local setup needed — you can run everything in the browser:

1. Go to the repository on GitHub
2. Click the green Code button → Codespaces tab → Create codespace on main
3. Once the environment loads, open workshop.ipynb and start coding

Workshop Outline

1. Getting the Data

Pull 5 years of SPY closing prices from Yahoo Finance using yfinance.

2. Building Signals (Bollinger Bands / Z-Score)

Compute a rolling SMA and standard deviation, then calculate a Z-Score to measure how far the current price deviates from its mean.

3. Generating Trades

Use np.where() to create trade signals:

• Long (+1) when Z-Score < −2 (oversold)
• Short (−1) when Z-Score > 2 (overbought)
• Flat (0) otherwise

Shift signals forward by one day to avoid look-ahead bias.

4. Vectorized Backtesting

Extract data to raw NumPy arrays and compute:

• Log returns for the market
• Strategy returns = position × market returns
• Cumulative returns via np.exp(np.cumsum(...))
• Annualized Sharpe Ratio

5. Visualizing Results

Plot the strategy equity curve against a buy-and-hold SPY benchmark.

Dependencies

Package	Purpose
pandas	Data manipulation & rolling statistics
numpy	Vectorized math & backtesting
matplotlib	Plotting equity curves
yfinance	Fetching historical market data

License

This project is maintained by the Rutgers Quant Finance Club.