Research note: This material is for educational and strategy-design purposes. It is not investment advice or a promise of future returns.
Executive summary
Pairs trading is a relative-value strategy that buys one stock and sells a closely related stock when their spread diverges, expecting convergence. The classic evidence comes from Gatev, Goetzmann and Rouwenhorst, and the strategy remains one of the best-known stat-arb templates.
In India, pairs trading works best among liquid same-sector names with comparable business models, such as private banks, large IT names, paints, cement or telecom-adjacent firms. Short-leg implementation, however, is decisive: if you do not have reliable shorting via futures or SLBS, the strategy becomes much harder to run cleanly.
Description
The main idea is to identify two stocks whose prices or log prices move together over time, then trade mean reversion of the spread. A simple distance method is easy to implement; a cointegration approach is statistically cleaner if the universe is large enough and the desk can maintain a stable calibration process.
For Indian books, same-sector and same-factor exposure pairs are safer than arbitrary statistical matches. You want business similarity, not just historical coincidence. That lowers model risk when regimes shift.
Key attribute table
The range below is a practical inference from pairs-trading literature and modern replication evidence.
| Time horizon | Turnover | Typical Sharpe or return expectation | Data needs | Complexity | |---|---|---|---|---| | Short to medium term | High | Sharpe roughly 0.8–1.5 gross; annualised return target roughly 6–12% after financing and borrow frictions | Daily prices, borrow or derivatives access, spread model | High |
Details
Universe: liquid same-sector NSE names with either stock futures or feasible borrow. Calibrate pair selection using a 6–12 month formation window. Entry when z-score of the spread exceeds 2.0; exit when it mean-reverts below 0.5 or after a maximum holding period, say 20–30 trading days. Size each leg to beta or volatility neutrality, not equal rupees by default. SLBS mechanics and equity-derivative access determine whether the short book is operational.
Risk controls: sector-neutral pairs only, fundamental-relationship check, avoid result-week entries, and hard stop if the spread widens beyond 3.5 standard deviations or a structural thesis break occurs. Borrow fees, SLBS availability and corporate-action adjustments must be modelled explicitly. NSE notes that corporate actions in SLBS can lead to foreclosure or adjustment, which is a real operational risk.
Backtests should compare distance and cointegration methods, test rolling stability of hedge ratios and simulate borrow unavailability. Edge cases include M&A rumours, sudden asset-quality developments in banks and sector-wide regime shifts that permanently change the relationship.
flowchart TD
A[Candidate liquid same-sector pair] --> B[Estimate hedge ratio]
B --> C[Compute spread z-score]
C --> D{Absolute z-score above two?}
D -->|No| E[No trade]
D -->|Yes| F[Enter long cheap leg short rich leg]
F --> G[Monitor borrow corporate actions and stop]
G --> H[Exit at mean reversion or time stop]
Implementation guide
- Select liquid same-sector stock pairs with economic similarity.
- Estimate hedge ratios and spread behaviour over a rolling window.
- Enter only when spreads move far enough from equilibrium.
- Neutralise beta or volatility between legs.
- Check short availability, borrow cost and event risk before trading.
- Exit on convergence, hard stop or thesis break.
India-specific example
A classic Indian candidate pair is HDFCBANK versus ICICIBANK, or INFY versus TCS. Suppose the log-price spread between INFY and TCS has a 250-day mean of 0.12 and a standard deviation of 0.04. If the current spread jumps to 0.21, the z-score is 2.25, which triggers a trade. If the hedge ratio is 0.9, you might buy ₹90 lakh of INFY and short ₹100 lakh equivalent of TCS, targeting spread convergence rather than outright market direction. The example relies on liquid large caps because that is where India’s shorting and execution infrastructure is strongest.
If the pair converges back to a z-score of 0.4 over eight sessions, the trade is closed. If TCS announces a major restructuring while INFY does not, the spread relationship may have structurally changed and the model should force an early exit. Corporate-action and result filters are as important as the statistical trigger.