Alpaca Quantitative Finance Script

Image of fulfilled orders
Image of portfolio holdings
Live stock tracking (AAPL in this example)

Inspiration

We were passionate about algorithmic trading and aimed to create a practical tool using real market data and quantitative strategies. After exploring statistical arbitrage and observing how quant firms leverage math-based models to exploit market inefficiencies, we decided to develop our own trading script with API integration and a functional backtester.

What It Does

The script identifies pairs of stocks with strong historical correlation and simulates trades when their price ratio deviates significantly from the mean—a classic statistical arbitrage strategy.

Key Functionalities:

Fetches historical and live price data
Identifies co-integrated stock pairs
Simulates trade entry/exit based on z-score thresholds
Visualizes trading signals, executed trades, and performance metrics
Tracks simulated portfolio value over time

How We Built It

We began by analyzing public GitHub repositories and existing trading models to understand their structure. Initially, we built on these models, then designed our own modular framework using:

Alpaca API for brokerage account access and real-time trade data
yfinance for historical stock price data
pandas and NumPy for time series data manipulation
matplotlib and Plotly for visualization and interactive dashboards

Our early backtester produced unreliable results, prompting us to pause and study statistical arbitrage concepts like cointegration, mean reversion, and z-score triggers. We then rewrote the core logic from scratch to ensure accuracy.

Challenges We Ran Into

The initial backtester gave misleading results due to flawed market execution assumptions
API rate limits required efficient caching and request throttling
Grasping statistical concepts like spread stability and stationarity was time-consuming
Implementing real-time visualization and strategy monitoring posed technical challenges

Accomplishments That We're Proud Of

Developed a working statistical arbitrage model that simulates realistic trades
Successfully integrated multiple APIs for historical and live data
Built a reliable backtester from scratch
Created clear visualizations to monitor strategy performance
Gained a deep understanding of quantitative trading mechanics

What We Learned

How to integrate financial APIs (Alpaca, yfinance) for real-time and historical market data
Fundamentals of statistical arbitrage, including mean reversion and cointegration
How to structure, test, and debug a trading strategy
The importance of effective visualizations for strategy analysis and iteration
Techniques for building and refining a backtester to mimic real-world trading