Pendulum - Itaú Quant Challenge

Welcome to Pendulum, a quantitative finance project developed for the Itaú Quant Challenge.

Overview

Pendulum focuses on executing a robust Statistical Arbitrage (Pairs Trading) strategy. The system is designed to dynamically screen the Brazilian equity market to identify cointegrated pairs, execute intraday and multi-day mean reversion trades, and rigorously manage portfolio risk.

Key Features

• Cointegration Testing: Utilizes Engle-Granger approach (Augmented Dickey-Fuller test) to statistically validate pair relationships.
• Statistical Arbitrage: Identifies temporary pricing anomalies between historically cointegrated assets (Pairs Trading).
• Backtesting Engine: Contains an institutional-grade Walk-Forward out-of-sample backtesting framework to validate alpha over time and reduce look-ahead biases.
• Risk Metrics: Tracks critical performance metrics such as Sharpe Ratio, Maximum Drawdown, Win Rate, and Gross Exposure, ensuring safe capital allocation.

Stack & Methodology

• Language: Python
• Core Libraries: Pandas, NumPy, Statsmodels, SciPy
• Data Source: B3 Historical Data (COTAHIST) and Yahoo Finance

Challenge Objectives

This project was structured to fulfill the rigorous criteria of the Itaú Quant Challenge, showcasing advanced techniques in quantitative research, backtesting realism, statistical validation, and risk control.