Data & analysis / Research prototype

Stat-Arb Engine

A Python research prototype for tracing a pairs-trading backtest from CSV inputs through signals, costs, and a report.

Python
Pandas
NumPy
Statsmodels

01 / The problem

The problem

A backtest needs enough detail to explain how it reached a result. This project keeps the input data, signals, position sizes, and costs available for inspection.

02 / What was built

What was built

CSV loaders align two price series by date. Separate modules calculate the hedge ratio, spread, z-score, position weights, and turnover costs. The report summarizes returns and drawdown.

03 / Engineering decisions

Engineering decisions

Match the dates

The loader sorts records, drops missing and duplicate rows, and keeps dates shared by both price series.

Return the working values

Backtest output includes prices, hedge ratio, spread, z-score, signals, weights, costs, returns, and equity in one table.

Keep costs adjustable

The cost module multiplies turnover by a configurable basis-point rate. A separate report calculates return, volatility, Sharpe ratio, drawdown, and total costs.

04 / Scope

Scope

As a researcher, I can load two CSV price series and align their dates.
As a developer, I can inspect signals, weights, and costs alongside the calculated returns.
As a researcher, I can change model settings and compare the output.

05 / Read the code

Source code

Reviewed September 10, 2026.