Pair Trading Strategy on NSE: Correlation, Ratio Signals and When the Pair Breaks

Most trading ideas ask one question: is this stock going up or down? A pair trading strategy on NSE asks a different one. Is this stock cheap or expensive compared with its twin? You buy one, sell the other, and the bet is on the gap between them, not on where the whole market is headed on Monday morning. When it works, it can pay even on a day the Nifty falls. When the relationship between the two stocks quietly breaks, it can hurt on both legs at once.

We’ll go through how traders pick a pair, the three ways to track it (spread, differential and ratio), how to turn the ratio into a standard-deviation reading you can act on, and a full worked example with two made-up tyre companies. Then the part most explainers skip: how to tell when a pair has stopped being a pair.

In this article

1. What a pair trading strategy on NSE actually bets on
2. Step one: does the pair actually move together?
3. Step two: pick what you’ll track
4. Step three: turn the ratio into a signal
5. A worked example: Malabar Tyres vs Konkan Tyres
6. When the pair breaks
7. The costs nobody puts in the example
8. A checklist before you put on a pair
9. The short version

What a pair trading strategy on NSE actually bets on

Start with two businesses that look alike in every way that matters. Zerodha Varsity’s Trading Systems module uses HDFC Bank and ICICI Bank as its opening illustration: both private sector banks, similar products, similar customers, the same regulator. If the RBI moves rates, both feel it in roughly the same way. So, all else equal, their share prices should move in a similar direction.

Now suppose one of them reports quarterly results, or a senior executive quits, or one stock gets caught in a burst of speculation. Varsity calls this a local event, something that hits one company in the pair and not the other. For a while the two prices drift apart. The pair trader’s view is that the drift is temporary. So you buy the one that has fallen behind and sell the one that has run ahead, and you wait for them to come back together.

That is the whole idea. Everything else, the correlation checks, the ratios, the standard deviations, is just a way of measuring “fallen behind” and “run ahead” without guessing.

A short history, and a disagreement

Investopedia’s pairs trading guide traces the strategy to technical analysts at Morgan Stanley in the mid-1980s. Varsity tells a similar story with names attached: it credits a Morgan Stanley trader called Gerry Bamberger with the first pair trade in the early 1980s, and another Morgan Stanley trader, Nunzio Tartaglia, with popularising it. It also notes that the hedge fund DE Shaw used the strategy in its early days.

Here’s where the two sources part ways. Investopedia describes pairs trading as a market-neutral strategy, because you’re long one stock and short another. Varsity pushes back hard on that label. Its argument is simple: you are long and short two different stocks, so you are not truly neutral. A genuinely neutral position, in Varsity’s view, is long and short the same underlying, the way a calendar spread is. It prefers to call pair trading “relative value trading”, and says some people call it statistical arbitrage.

We think Varsity has the better of this argument, and it isn’t just semantics. If you believe you’ve built a risk-free hedge, you’ll size it too big. Treat it as two separate risks that usually offset each other, and you’ll size it sensibly.

A pair trade is not a hedge that removes risk. It is two positions whose risks usually cancel, and “usually” is the word that costs money when it fails.

Step one: does the pair actually move together?

Being in the same sector isn’t enough. Varsity asks this question directly: do ICICI Bank and HDFC Bank qualify as a pair just because both are private banks? Its answer is to measure it, using correlation.

Correlation runs from -1 to +1. A reading close to +1 means the two series tend to move in the same direction; zero means no relationship. Two details matter for traders. First, correlation tells you about direction, not size. A +0.75 correlation doesn’t mean that when stock A rises 3%, stock B also rises 3%. Second, the correlation of A against B is the same as B against A, so it doesn’t matter which one you put first.

How high is high enough? The two sources give slightly different bars. Varsity says that in this method the correlation number should ideally be above 0.75. Investopedia says pairs need a high correlation, typically above 0.80, and admits that such pairs are hard to find.

Varsity’s own test is a useful reality check. It took Axis Bank and ICICI Bank, downloaded roughly two years of closing prices (4 December 2015 to 4 December 2017, about 496 data points), and found a correlation of just 0.51 on closing prices. Two private banks, similar businesses, and the number came in well under its own 0.75 bar. Intuition said “pair”; the data said “maybe”.

Getting the data right

Varsity flags two data rules before you calculate anything:

• Both stocks need the same number of data points, matched to the same dates. If one stock was suspended for a week, line the dates up rather than letting the series slide out of step.
• Clean the prices for corporate actions such as bonuses and splits. An unadjusted 1:1 bonus halves the price overnight and wrecks every ratio after that date.

It also suggests calculating correlation on daily returns, not just closing prices, where the daily return is today’s close divided by yesterday’s close, minus 1. Varsity says it isn’t a big fan of correlation on raw closing prices, and that’s sensible: two stocks that both trend up over two years will show a decent price correlation even if their day-to-day moves have little in common.

Step two: pick what you’ll track

Once you have a candidate pair, you need one number that sums up the relationship each day. Varsity’s module walks through three:

• Spread. In Varsity’s correlation method, this is the difference between the two stocks’ daily moves, each measured against its previous close. If stock 1’s move reads 6.1 and stock 2’s reads 3.85, the spread is 2.25. If stock 2 had instead moved -3.85, the spread would be 9.95.
• Differential. The plain difference between the two closing prices. A stock at ₹175 and another at ₹232 gives a differential of -57. Varsity says this works best on an end-of-day basis, not intraday.
• Ratio. Price of stock 1 divided by price of stock 2. Varsity notes the ratio tends to look more consistent as a time series than the other two, and it’s the one its trade rules are built on.

We’ll use the ratio from here on. Whichever stock is on top (the numerator) decides which leg you buy and which you sell, so write it down and don’t switch halfway through.

Step three: turn the ratio into a signal

A ratio of 1.73 means nothing on its own. What matters is how far it sits from its own normal level. So for the look-back window you’ve chosen, calculate two things: the mean (average) of the ratio and its standard deviation. Then, for today:

Distance from the mean in standard deviations = (today’s ratio – mean ratio) ÷ standard deviation of the ratio

This is what traders usually call a z-score. A reading of +2.5 means today’s ratio is two and a half standard deviations above its average. Varsity gets to the same place a different way: it converts the ratio into a “density curve” value and reads its trade triggers from that. A density value between 0.025 and 0.003 corresponds to a ratio between two and three standard deviations below the mean; between 0.975 and 0.997 corresponds to two to three standard deviations above.

Varsity’s guideline for trading this:

• Enter when the ratio is between the second and third standard deviation from the mean.
• Long pair trade (ratio unusually low): buy the numerator, sell the denominator.
• Short pair trade (ratio unusually high): sell the numerator, buy the denominator.
• Exit as the ratio moves back towards the mean.
• Stop-loss pegged at the third standard deviation in its guideline table.

Varsity’s memory trick is handy: whatever the trade says about the pair, do it to the numerator, and do the opposite to the denominator.

A worked example: Malabar Tyres vs Konkan Tyres

Both companies here are invented and every price is hypothetical, picked to keep the arithmetic clean. Imagine two mid-sized tyre makers in the F&O segment. Same raw materials, same customers, same exposure to rubber prices. Over the last 250 sessions, their daily returns show a correlation of 0.82, which clears both the Varsity and Investopedia bars.

We define the ratio as Malabar ÷ Konkan. Over that window:

• Mean ratio: 1.60
• Standard deviation of the ratio: 0.05

So the two-standard-deviation band runs from 1.50 to 1.70, and the three-standard-deviation band from 1.45 to 1.75.

The setup

Malabar posts a strong quarter and the stock jumps. A few sessions later it closes at ₹512, while Konkan sits at ₹296.

Ratio = 512 ÷ 296 = 1.7297. Distance from the mean = (1.7297 – 1.60) ÷ 0.05 = 2.59 standard deviations.

That sits between the second and third standard deviation, on the high side. Malabar looks expensive relative to Konkan. It’s a short pair trade: sell Malabar (the numerator), buy Konkan (the denominator).

Balancing the two legs

Varsity’s own example runs both legs in stock futures, and that’s the practical route here, since a pair trade can stay open for a long time. It also points out that ideally you want the same rupee value on the long and short sides, which it calls rupee neutrality. Lot sizes rarely line up neatly, so you get as close as you can.

Say (hypothetically) Malabar’s futures lot is 1,000 shares and Konkan’s is 1,700.

• Short 1 lot Malabar: 1,000 × ₹512 = ₹5,12,000
• Long 1 lot Konkan: 1,700 × ₹296 = ₹5,03,200

The gap is ₹8,800, under 2% of either leg. Close enough. If one side had been double the other, you’d be running a directional bet with a small hedge attached, not a pair trade.

Outcome A: the pair converges

Three weeks later the broader market has a bad patch and both stocks fall. Malabar drops to ₹488 and Konkan slips to ₹301.

New ratio = 488 ÷ 301 = 1.6213, back near the 1.60 mean. Time to exit.

• Short Malabar: (512 – 488) × 1,000 = +₹24,000
• Long Konkan: (301 – 296) × 1,700 = +₹8,500
• Total, before brokerage, taxes and other charges: +₹32,500

Look at where the money came from. Most of it was the short leg, because Malabar, the stock that had run ahead, gave back its extra gain. Varsity makes the same observation from its real examples: the bulk of the profit usually comes from whichever stock had wandered away. And notice that the market fell. A straight long position in either stock would have lost money on Malabar.

Outcome B: the pair keeps stretching

Now the other version. Malabar’s results weren’t a blip. Analysts start upgrading it and the stock climbs to ₹530. Konkan edges up to ₹302.50.

Ratio = 530 ÷ 302.50 = 1.7521. That’s past the third standard deviation at 1.75. The stop is hit.

• Short Malabar: (512 – 530) × 1,000 = -₹18,000
• Long Konkan: (302.50 – 296) × 1,700 = +₹11,050
• Total, before charges: -₹6,950

The Konkan leg cushioned the damage, which is the point of having two legs. Measured against the ₹32,500 you’d make in Outcome A, the loss at the stop is small. That asymmetry only holds if you actually take the stop.

The stop on a pair trade isn’t about price. It’s about the moment the ratio tells you the two stocks may no longer belong together.

When the pair breaks

This is the real risk, and it’s worth spending time on. The whole strategy rests on the idea that the relationship is stable and the gap is noise. Investopedia’s warning is blunt: past prices are not always a guide to the future, and a correlation that held for years can falter.

Varsity’s framework gives a clean way to think about it. A local event, like one quarter’s results or a sudden bout of speculation, is a temporary disturbance. The pair snaps back. But some events change the business itself: a big acquisition, a new product line, a change in management, a regulatory hit that lands on one company and not the other. After that, the old mean ratio may not be the right mean any more. You’re waiting for a reversion that isn’t coming.

A few practical warning signs:

• The correlation is sliding. Recalculate it on a rolling window. If a pair that sat above 0.80 is now drifting towards 0.6, the past mean ratio means less.
• The ratio is trending, not swinging. Varsity notes that correlation analysis only makes sense when the data is “stationary around the mean”, meaning it keeps returning to its average. A ratio that steps up and stays up is telling you the mean has moved.
• The news is about the business, not the quarter. One weak result is noise. A merger, a lost licence or a new major shareholder is structure.
• The trade is old and going nowhere. Varsity says a pair trade can stay live for an extended period. That’s fine if the ratio is drifting back. If it’s stuck at the edge for weeks, that’s a question worth asking.

This is also why the Varsity module doesn’t stop at correlation. Its later chapters move on to linear regression and the ADF test, a statistical check for whether a series is stationary. That’s a more rigorous way of asking “does this pair actually revert?”, and it’s worth learning once the basic version makes sense.

The costs nobody puts in the example

Our worked example ignored charges. Real ones won’t. You’re paying brokerage and taxes on four transactions (two to open, two to close), and on futures you’re posting margin on both legs. Varsity calls pair trading a “margin money guzzler” and says you need sufficient funds to do it.

There’s also the waiting. Varsity’s estimate is that a single pair may throw up only two or three signals a year, so traders need to watch several pairs to find steady opportunities. That’s a lot of spreadsheet work for occasional trades. Anyone expecting daily action from one pair will either get bored or start forcing trades at 1.5 standard deviations, which is how the edge disappears.

Position sizing matters more here than it looks. Size the trade off the loss at your stop (the ₹6,950 in Outcome B), not off the combined contract value. Our position sizing formula guide shows how to work backwards from a fixed rupee risk, and the broader rules are in our risk management in trading guide. If you’ve traded single-stock reversion before, our piece on mean reversion trading covers the same instinct applied to one chart instead of two.

A checklist before you put on a pair

• Is there a real business reason the two stocks should move together, beyond sharing a sector label?
• Does the correlation of daily returns clear your bar (Varsity’s 0.75 or Investopedia’s 0.80) on a reasonably long window?
• Are both price series matched by date and adjusted for bonuses and splits?
• Is the ratio between two and three standard deviations from its mean, and is it swinging rather than trending?
• Is there a structural reason for the gap (merger, management change, regulatory action) rather than a one-off event?
• Are the two legs close in rupee value?
• Do you know your exit at the mean and your stop at the third standard deviation, and the rupee loss if the stop is hit?
• Do you have margin for both legs, with room to spare if the trade sits open for weeks?

The short version

Pair trading means buying one stock and selling a close cousin when the price relationship between them stretches further than usual, then exiting when it settles back. Check the pair with correlation on clean, date-matched data, track the price ratio, and measure how many standard deviations it sits from its mean. Varsity’s rule of thumb is to enter between the second and third standard deviation, exit near the mean and stop out at the third. The biggest risk isn’t a noisy day; it’s a pair that has quietly stopped being a pair. Size for that, and don’t mistake two offsetting positions for no risk at all.

Disclaimer: This article is for information and education only and is not investment advice or a recommendation to buy or sell any security. Malabar Tyres, Konkan Tyres and all prices, lot sizes, correlations and ratios in the worked example are hypothetical. Pair trading definitions, thresholds and trade rules are from the Zerodha Varsity and Investopedia pages linked above. Derivatives trading involves substantial risk, including losses beyond the margin posted; consult a SEBI-registered investment adviser before making investment decisions.