LongandShortR: Statistical Arbitrage & Cointegration Engine on B3 in R

Organization / Context: Meta & Actio Investimentos
Timeline: 2017 - 2020
Domain Category: Quantitative Finance
Technologies: R, Tseries, Urca, Dplyr, Ggplot2, B3 Market Data

Executive Overview

Systematic quantitative algorithm coded in R for daily universe-wide Engle-Granger cointegration and Augmented Dickey-Fuller (ADF) screening on B3 equities. Scanned 100+ statistical arbitrage candidates monthly with automated half-life and z-score calculations.

Architecture & Data Flow

The diagram below illustrates the sanitized high-level component topology and data processing pipeline:

flowchart TD
  A[Dados de Cotações Históricas B3] --> B[Testes de Raiz Unitária ADF]
  B --> C[Regressão e Cointegração Engle-Granger]
  C --> D[Cálculo de Meia-Vida e Z-Scores]
  D --> E[Geração de Sinais de Arbitragem]
  E --> F[Painel Analítico de Pares Elegíveis]

Key Engineering Highlights

• Systematic B3 Screening: Automated daily Augmented Dickey-Fuller (ADF) and Engle-Granger cointegration testing across liquid equity pairs.
• Quantitative Signal Generation: Residual spread modeling with mean-reversion half-life calculation and Z-score standardization.
• Execution Risk Framework: Entry/exit band sizing calibrated for transaction costs and statistical confidence intervals.