Pair trading is a market-neutral strategy that goes long one security and shorts another to profit from their relative mispricing. In India, it can be implemented using equities, futures, or SLB (stock lending & borrowing), but success depends on robust statistical testing, disciplined risk controls, and awareness of SEBI/NSE rules.
Why pair trading matters in India
Indian stocks within the same sector (e.g., large banks or IT majors) often move together due to shared fundamentals. When two such securities temporarily diverge, a trader can buy the underperformer and sell the outperformer, betting that the “spread” will revert. Decades of academic work show that well-designed pairs strategies can earn excess returns, especially when rigorously backtested and executed with low costs. (www-stat.wharton.upenn.edu)
How pair trading works
- Find a relationship: Identify two securities with a stable long-run link (ideally, cointegration, not just correlation).
- Build a spread: Combine prices so that the spread is stationary (mean-reverting).
- Trade deviations: When the spread moves far from its mean (entry signal), go long/short the pair; exit when it reverts.
Minimal math you’ll use
- Hedge ratio (β) via OLS:
β=Cov(lnPA, lnPB)Var(lnPB)\beta = \frac{\mathrm{Cov}(\ln P_A,\ \ln P_B)}{\mathrm{Var}(\ln P_B)}
- Spread: st=lnPA−βlnPBs_t = \ln P_A – \beta \ln P_B
- Z-score signal: zt=st−μsσsz_t = \frac{s_t – \mu_s}{\sigma_s}
- Typical rules: enter at ∣z∣≥2|z|\ge 2, exit near z=0z=0.
- Half-life of mean reversion (AR(1) on sts_t):
Fit Δst=ϕst−1+εt \Delta s_t = \phi s_{t-1} + \varepsilon_t. Half-life ≈ ln(0.5)/ln(1+ϕ)\ln(0.5)/\ln(1+\phi).
Correlation vs cointegration (at a glance)
- Correlation = move together today; can drift apart over time.
- Cointegration = a stable long-run equilibrium; deviations are mean-reverting. Use Engle-Granger (2-step) or Johansen tests to verify. (Federal Reserve, IMF)
Instruments Indian traders actually use
- Futures–Futures (same expiry): Common for NIFTY50 stocks or within a sector (e.g., Bank A vs Bank B). Clean shorting, exchange-cleared, straightforward margins.
- Cash–Futures: Long cash, short futures (or vice-versa) to express relative value when one leg lacks liquidity in derivatives.
- Equity–ETF (e.g., largecap stock vs NIFTYBEES): To play stock vs basket dislocations (watch tracking error/fees).
- Cash–Cash via SLB: Short cash stock through the Securities Lending & Borrowing (SLB) framework on NSE; mind borrow availability, fees, and recall risk. (NSE India)
Practical, India-centric workflow
1) Universe & data
- Start with F&O-eligible large/mid caps for liquidity.
- Use at least 2–3 years of clean daily (or higher-freq) data; adjust for splits, bonuses, demergers, and index rebalances.
2) Pair selection (avoid data-mining)
- Economic rationale: same sub-sector (e.g., TCS–Infosys, HDFC Bank–ICICI Bank, Hindalco–NALCO)—examples only; test before trading.
- Stat tests:
- Check individual series are I(1), then test cointegration (Engle-Granger/Johansen).
- Validate out-of-sample; reject pairs whose cointegration breaks post-event. (Federal Reserve, IMF)
3) Entry/exit design
- Compute rolling β; form spread and z-score.
- Entry: ∣z∣≥2|z|\ge 2 (tune to your slippage/costs).
- Exit: z→0z\to 0 or take-profit at ∣z∣≤0.5|z|\le 0.5; stop-loss at ∣z∣≥3|z|\ge 3 or time-stop = 5–10 trading days.
- Position sizing: Target a rupee-neutral or beta-neutral book so that index moves wash out.
4) Risk controls that move the needle
- Corporate-action filter: Skip around results, mergers, demergers, block deals.
- Volatility gates: Avoid entries when IV percentile is extreme or on macro event days (RBI policy, Union Budget).
- Ban-list check: If a stock enters F&O ban (OI ≥ 95% MWPL), you may be unable to build/modify positions; design contingencies. (NSE Clearing)
- Margin readiness: SEBI’s framework on upfront/peak margin and ongoing enhancements to derivatives risk monitoring require sufficient capital buffers and clean margin reporting. (Securities and Exchange Board of India, Reuters)
Opportunities unique to the Indian market
- Sector-heavy dispersion: Indian sectors often see transient divergences (policy changes, promoter actions), creating rich mean-reversion windows—especially in private banks, IT services, metals, and autos.
- Event-driven spreads: Index rebalances, MSCI flows, and corporate actions can push pairs off-kilter, then normalize.
- Cash-futures mispricings: Basis swings (dividends, funding) can widen spreads briefly—ripe for relative trades if hedged tightly.
Evidence from classic research indicates that simple distance/cointegration-based pairs can be profitable after costs when implemented systematically—use this as a starting point and adapt to India’s frictions and rules. (www-stat.wharton.upenn.edu)
Risks (and how to manage them)
- Relationship breakdown
- Risk: Structural shifts (management change, regulation, capex cycles) can kill cointegration.
- Mitigation: Retest cointegration monthly/quarterly; drop pairs after repeated stop-outs or regime change. (Federal Reserve)
- Liquidity & execution
- Risk: Wide spreads, partial fills, and slippage—worse in midcaps and during volatile opens/closes.
- Mitigation: Use futures where possible; schedule entries outside auction extremes; pre-trade impact checks.
- Borrow & recall (for cash shorts)
- Risk: SLB fees spike; borrows can be recalled.
- Mitigation: Prefer futures–futures; if using SLB, track fee/availability dashboards and recall notifications. (NSE India)
- Regulatory constraints
- Risk: F&O ban halts fresh positions; margin/risk-monitoring rules can force deleveraging and penalize breaches.
- Mitigation: Integrate ban-list and margin headroom checks in your OMS; avoid expiry-day crowding. (NSE Clearing, Securities and Exchange Board of India, Reuters)
- Overfitting & look-ahead bias
- Risk: Data-mined pairs look great in backtests but fail live.
- Mitigation: Use walk-forward validation, transaction-cost models, and out-of-sample filters.
A simple implementation blueprint
Step 1: Data hygiene
- Corporate-action-adjusted close prices; align trading calendars; remove bad ticks.
Step 2: Pair formation
- For each sector, pre-screen by correlation > 0.7; then run Engle-Granger/Johansen to confirm cointegration.
- Keep pairs with stable β and residual stationarity.
Step 3: Signal engine
- Rolling 120-day β; compute spread and z-score.
- Enter: ∣z∣≥2|z|\ge 2; Exit: z→0z\to 0; Stop: ∣z∣≥3|z|\ge 3 or 10-day timeout.
- Position size: Choose lots so that ₹ exposure (or β) is neutral; cap per-pair risk at 0.5–1.0% of equity.
Step 4: Risk overlays
- Event calendar (RBI, earnings) → block new entries T-1 to T+1.
- Volatility filter (e.g., IVP 80+ or intraday gap > 1.5% → skip).
- Ban-list & margin checks before order release.
Step 5: Execution
- Use limit-at-touch or algo slices (VWAP/TWAP) for the leg with poorer liquidity; cross the spread on the tighter leg.
- Monitor basis on futures and expected dividend differentials.
Featured table: Where to express a pair
| Method | Pros | Cons | When to use |
| Futures–Futures | Easy shorting, cleaner margins, low borrow risk | Roll/expiry management | Large-cap, liquid pairs |
| Cash–Futures | Flexibility if one leg lacks F&O | Basis/dividend complexities | One leg illiquid in derivatives |
| Cash–Cash via SLB | Access to non-F&O names | Borrow fees & recall | Longer-horizon trades with stable borrows |
Note: Check F&O ban (MWPL ≥ 95%) before trading; positions may only be reduced during ban. (NSE Clearing)
Compliance corner (what to know before you deploy)
- Upfront/peak margin verification is mandatory across cash and derivatives; brokers/clearers validate peak intraday margins. Ensure funding buffers and accurate client-level reporting. (Securities and Exchange Board of India)
- Ongoing tightening of derivatives risk oversight (e.g., curbs on contract proliferation and enhanced monitoring) affects sizing, expiries, and intraday risk. Build flexibility into rules and dashboards. (Reuters)
FAQ (quick answers)
Is high correlation enough?
No. Two prices can be highly correlated yet drift apart permanently. You want cointegration—a statistically stable long-run equilibrium. (Federal Reserve)
What sample size is “enough” to test cointegration?
Johansen tests are asymptotic; larger samples are more reliable. In practice, use multi-year daily data and validate out-of-sample. (IMF)
What if one leg hits F&O ban mid-trade?
You can usually only reduce positions, not increase or initiate fresh ones. Code a contingency (e.g., flatten both legs). (NSE Clearing)
Can I short cash equities reliably in India?
Yes, via SLB on NSE, subject to borrow availability, fees, and recall. Many practitioners prefer futures for simplicity. (NSE India)
Key takeaways for Indian investors
- Pair trading can extract alpha from temporary dislocations while staying largely market-neutral.
- The edge lies in sound statistics, disciplined exits, and friction-aware execution (margins, ban lists, SLB).
- Treat regulation and microstructure as first-class inputs to your model—not afterthoughts.