UK Pair Trading Strategy

Pair trading involves going long one stock and short another related stock. You profit from the spread between them converging to its mean, regardless of market direction. It's market-neutral because you're hedged.

Z-score measures how far the spread is from its mean in standard deviations. Z = -2 means spread is 2 StdDevs below mean (oversold, go long spread). Z = +2 means 2 StdDevs above (overbought, short spread).

Best pairs have high correlation (>0.7), pass cointegration test, and are in the same sector. UK examples: Shell/BP (energy), Lloyds/Barclays (banks), Rio/Anglo (mining), GSK/AZN (pharma).

Having equal dollar amounts long and short eliminates market exposure. S$10,000 long Shell + S$10,000 short BP means market moves don't affect you, only relative performance matters.

Exit when: (1) Z-score returns to ~0.5 or 0 (target), (2) Z-score expands to ±3.5 (stop loss), or (3) 20 trading days pass (time stop). Also exit on major fundamental news.

Use OLS regression: regress Stock A on Stock B, the slope is your hedge ratio. For $10,000 long A with hedge ratio 1.15, short $11,500 of B. This accounts for different volatilities.

Cointegration (ADF test) shows the spread has a long-term equilibrium. Unlike correlation (short-term), cointegration ensures spread won't drift apart permanently. Need p-value < 0.05.

UK stamp duty (0.5%) + commissions + spreads = ~0.9% round-trip. Significant for pair trading. Use CFDs for short leg to avoid stamp duty. Net returns must exceed costs.

4-8 pairs from different sectors provides diversification. Equal risk allocation (2% each). Different sectors (energy, banks, staples, utilities) reduces correlation between pairs.

Half-life measures how fast spread reverts to mean. Shell/BP ~10 days, Banks ~7 days. Shorter is better. Use to set holding period expectations and time stops.

Kalman filter provides dynamic hedge ratio that adapts in real-time as relationship changes. More responsive than static OLS. State-space model estimates unobserved hedge ratio from spread.

OU process models mean-reverting behavior with parameters: mean level, reversion speed, volatility. Half-life = ln(2)/speed. Used to estimate expected convergence time and optimal parameters.

ML applications: (1) Pair selection - predict profitable pairs, (2) Spread prediction - forecast Z-score, (3) Regime detection - identify correlation states. Challenge: alpha decay is fast.

Trade implied vs realized correlation. Buy straddles on individual stocks, sell index straddle. Profit if actual correlation lower than implied (stocks diverge more). Related to pair trading.

Break down: by pair (contribution), by sector (unintended exposure), transaction costs (drag), timing alpha (entry/exit skill), market residual (neutrality leakage). Target Sharpe >1.0, win rate >55%.