PYTHONadvanced~75 min

Pairs Trading & Cointegration

Estimate a hedge ratio, test the spread for stationarity and half-life, then backtest a z-score strategy honestly with costs and an out-of-sample split.

Task brief

README.md

# Pairs Trading & Cointegration

**Role relevance:** Quant research / statistical arbitrage take-home
**Estimated time:** 75 minutes
**Difficulty:** Advanced
**Format:** Jupyter notebook (.ipynb)

## What you are given
- pairs_trading_starter.ipynb - a guided notebook that runs end to end even before you fill it in
- prices.csv - 750 business days of two price series
- Validation targets in the final cell, so you can grade your own work

## What you must deliver
1. The hedge ratio, fitted on the in-sample window only
2. An ADF stationarity test and the spread's half-life
3. A causal z-score signal with the position lagged one day
4. A cost-aware backtest reporting in-sample and out-of-sample separately, plus a written verdict

## Constraints & assumptions
The hedge ratio is fitted on the first 500 rows only. The z-score must be rolling, not full-sample. Positions must be lagged before they earn PnL. Costs are 5bp on turnover.

## Submission note
Complete the starter file, then compare your work against the mark scheme.

What you'll learn

Tell cointegration apart from correlation, and test for it
Use the half-life to set a sensible holding period and thresholds
Separate a real edge from an overfit with costs and a holdout