Clustered pairs portfolio
This strategy runs many pair trades at once instead of one. It sorts a list of 36 large-cap F&O stocks into eight groups whose daily returns behave alike, picks the most closely related pairs inside each group, and trades the pairs whose price gap has stretched furthest from normal. Each pair is long one stock future and short the other, held until the gap settles.How it trades
Results
PRE-BUILT · DEFAULTSWhere it made and lost money
Although each trade is long one stock and short another, results in this test tracked the NIFTY backdrop closely, so the pairs were not fully market-neutral. Trades held while NIFTY rose over 1% made +₹2,59,279 over 14 trades; those held while it fell over 1% lost ₹1,54,536 over 13. Entry days when NIFTY closed below its open earned ₹3,29,976 over 21 trades, while days it closed above its open lost ₹2,97,162 over 25. Quiet entry sessions lost ₹1,04,763 over 26 trades and calm evenings with India VIX below 11.5 lost ₹77,680 over 12. Whole-lot rounding and stocks with different sensitivity to the index leave a net market tilt in each pair, and that tilt drove much of the outcome.
About this strategy
- Every 21 sessions, standardise each stock's last 250 daily returns and group the 36 stocks into 8 clusters of similar behaviour.
- Inside each cluster keep the 10 most correlated pairs and fit each pair's hedge ratio from log prices over the same 250 sessions.
- Each evening, measure every pair's spread as a z-score against its previous 60 closes.
- At 09:20 next session, open pairs whose z-score is beyond ±2.0, most stretched first: buy the cheap stock's future, sell the rich one. The first stock is 1 lot; the second is value-matched by the hedge ratio in whole lots. Both legs sit in one position.
- Only one open pair per cluster, no stock in two pairs, none in the F&O ban, and at most 15 pairs in all.
- After at least 5 sessions, close a pair at 09:20 once its z-score is back within ±0.5.
- On monthly expiry every open pair is closed at 15:15 and reopened in the next month at 09:20 the following session.
Why traders use it
A single pair gives few trades and depends on one relationship holding. Spreading the idea across a broad universe gives more opportunities and spreads the risk of any one link breaking. Clustering by return behaviour lets the data decide which stocks belong together, rather than relying on sector labels alone. Because each pair is long one stock and short another, the hope is that broad market swings cancel and the profit comes from relative moves.
The year in plain words
From 1 Oct 2025 to 30 Sep 2026 on default settings, the portfolio closed 46 trades and made ₹32,813 before costs, with a 52% win rate and a profit factor of 1.05. After ₹20 per order and bid-ask slippage it lost ₹11,255. Each monthly roll closes a pair and reopens it, so 27 of the 46 trades ended at a roll and only 17 on the spread normalising; the median hold was 18 days. Swings were large: February made ₹2,03,165 but January lost ₹1,54,328, July lost ₹87,720, and the drawdown reached ₹2,38,855 before costs. Six of 12 months were positive. Overall it roughly broke even.
What the settings do
Risks
- There is no stop: a spread that keeps widening stays open until it normalises or the test ends, and the largest single loss was ₹93,705 before costs.
- Whole-lot sizing and stocks with different index sensitivity leave each pair with a net market tilt, which drove much of the P&L here.
- Several pairs open at once need margin for many stock futures, and a drawdown above ₹2.3 lakh before costs is possible.
- Closing and reopening every pair at each monthly expiry adds four orders per pair, which turned a small gain into a loss after costs.
Who it suits
Experienced futures traders with substantial margin who want a rules-based relative-value portfolio, can monitor many positions and rolls, and can tolerate deep, lasting drawdowns for results close to break-even in this test.
Questions traders ask
Is clustered pairs trading profitable on NSE stock futures?
Barely, in this test. From 1 Oct 2025 to 30 Sep 2026 the default portfolio made ₹32,813 before costs but lost ₹11,255 after ₹20 per order and bid-ask slippage, with a drawdown of ₹2,38,855 before costs. Statutory charges are not included. Past results do not guarantee future ones.
What does clustering add to pairs trading?
Instead of choosing pairs by sector label, clustering groups stocks whose daily returns have actually moved alike over the past 250 sessions. Pairs are then picked only inside a group, and only one pair per group can be open, which spreads trades across different parts of the market.
Why did a market-neutral pairs strategy follow NIFTY?
The pairs are matched by value and hedge ratio, but lots must be whole and the two stocks rarely react to the index equally. Each pair therefore keeps some net long or short market exposure. In this test that leftover tilt mattered: trades held through a NIFTY rise over 1% made money and those held through a fall over 1% lost.
Why are there so many monthly roll exits?
Stock futures cannot be held past expiry, so every open pair is closed at 15:15 on expiry day and reopened the next morning. Each of those segments counts as a separate trade, which is why 27 of the 46 trades ended at a roll.