Stock Pairs Trading

No. Pairs trading requires short selling, which is not permitted in registered accounts (TFSA, RRSP). You need a margin account for pairs trading.

Minimum $25,000 is recommended. You need enough to take meaningful positions on both sides (long and short) while maintaining required margin. Some start with $10,000 but positions will be small.

Start with RY/TD (Royal Bank vs TD Bank). It's the most liquid, highest correlation pair with reliable mean reversion. Once comfortable, try other bank pairs or ENB/TRP, CNR/CP.

Typically 5-20 days for most pairs. The half-life of the spread (time to revert halfway) guides expectations. Some trades converge in days; others may take weeks. Use time stops (30 days max) if not converging.

If you're short when a stock goes ex-dividend, you owe the dividend. You pay the dividend to whoever you borrowed the shares from. Factor this into your P&L and be aware of ex-dividend dates.

Correlation measures how returns move together short-term but can break down. Cointegration tests for a long-term equilibrium relationship between prices, ensuring the spread is statistically mean-reverting. Cointegration is stronger for pairs trading.

Simple method: Price ratio or dollar-neutral. Better method: Regress Stock A prices on Stock B prices; the slope (beta) is the hedge ratio. Most accurate: Use rolling regression or Kalman filter for dynamic hedge ratio.

Exit when: 1) Z-score returns to 0 (or 0.5 for conservative), 2) Stop loss triggered (Z > 3.5), 3) Time stop reached (30 days), or 4) Correlation breakdown detected. Define exits before entering.

4-6 pairs is typical for good diversification without over-complexity. Ensure pairs are from different sectors. More capital allows more pairs. Quality over quantity - only trade pairs with strong relationships.

M&A (one company gets acquired), fundamental change (one company's business shifts), earnings surprise (dramatically different results), credit event, or sector rotation. When divergence is fundamental, the pair may not revert. Exit and reassess.

Fit an AR(1) model to the spread: Spread(t) = α + β × Spread(t-1) + ε. Half-life = -ln(2) / ln(β). Alternatively, from Ornstein-Uhlenbeck: Half-life = ln(2) / θ where θ is the mean-reversion speed parameter.

Kalman filter estimates hedge ratio and spread parameters dynamically, treating them as hidden states that evolve over time. It adapts to changing relationships better than static regression. More complex but more robust.

Use CUSUM test (cumulative sum of deviations), Bai-Perron test for multiple breaks, or monitor rolling correlation. Sudden correlation drops, cointegration test failures, or spread permanently shifting indicate structural breaks.

Screen universe for correlation/cointegration. Test multiple lookback periods, entry/exit Z-scores. Walk-forward optimization. Out-of-sample testing. Account for transaction costs and short borrow. Avoid overfitting.

Protective: Buy puts on short leg, calls on long leg for spread blowout protection. Synthetic: Use synthetic longs/shorts via options to reduce capital. Volatility: If IVs diverge, can add option spread. Complexity increases - start with stock-only pairs.