On 10 Jun 2026 the Nifty's at the money volatility for the 30 Jun expiry was 15.03%, and the fifty stocks inside it were priced higher on average. A dispersion trade sells the first and buys the second. By 30 Jun the index had realised less than it was priced at and the stocks about what they were priced at, so on paper the trade made money. Sold at the June levels and then met with the twenty session realised correlation of 0.60 and index realised volatility of 29.46 per cent of 8 Apr 2026, it would have lost. April's implied figures are not in the data, so that is a what if. This lesson builds the trade leg by leg, shows its profit and loss as a function of correlation, and spends as long on the way it fails as on the way it works.
3By the end you can
- Describe the legs of a dispersion position and how their vega is matched.
- Compute the profit of the index leg as a function of realised correlation.
- List the Indian constraints on the trade and the market draw that hurts it.
The idea, plainly
The trade in one sentence
Sell the volatility of an index. Buy the volatility of its stocks, in sizes that add up to the same vega. Hedge the delta of every leg. What is left is a bet that the stocks will move apart more than the market priced in. In other words, you are short correlation.
Dispersion is short index volatility and long stock volatility, hedged
Why it is a correlation trade
The formula of the last lesson ties the index volatility to the stock volatilities and one correlation. Hold the stock volatilities where the market priced them. Then the only thing that can change the index variance is the correlation, and the profit of the trade, in units of variance, is the gap between the correlation priced and the correlation realised, times a fixed number that depends on the weights and the volatilities.
Profit in variance =
(rho implied − rho realised)
x (A − B)
Module E07 defines A and B. If the realised correlation is lower than the implied one, the index realises less variance than it was sold at, and the short index leg gains. If it is higher, the leg loses.
The maths: profit as a function of correlation
Let the index variance be B plus rho times (A minus B), with the stock volatilities fixed. The index volatility that a realised correlation gives is the square root of that.
Index vol(rho) =
sqrt( B + rho x (A − B) )
The short index leg earns the implied index volatility minus the realised one, in volatility points, per unit of vega. Put the stock legs at zero profit and the whole trade earns that. Because of the square root, the profit is not a straight line in rho: it bends. The same distance of correlation is worth a little more in volatility points below the implied level than above it. The shape is not the danger. The danger is the size and the timing of the move. Realised correlation went from the low thirties to its peak within about a month in 2026, over a stretch in which the index fell sharply and then rebounded.
Vega weighted or gamma weighted
The weights of the stock legs can be set two ways. Vega weights make the position flat to a parallel move in all the volatilities at the start. Gamma weights make the day to day gains and losses from realised moves line up: the stock legs and the index leg then earn and lose together when the stocks and the index move by a given amount. The two agree when every option has the same expiry and is at the money, and drift apart as strikes and expiries differ. Say which one a result uses. A vega weighted position that is not rebalanced is a gamma weighted one only for the first day.
2 terms this section turns on
Building it
- Choose the basket. The heaviest names give most of the answer. Ten names held a bit over half of the index weight on 10 Jun 2026, and gave nearly the implied correlation of all fifty.
- Match the vega. Each stock's straddle gets a share of the index straddle's vega equal to its weight. Lots are whole numbers and differ by stock, so the match is never exact. Rounding is a leg of the trade.
- Use the same expiry. A mismatch of expiries adds a term structure bet to a correlation bet.
- Hedge delta. At the start and as the legs move, with futures in the index and in each stock. Every rebalance has a spread and a charge, and the paper result below leaves them out.
- Plan the exit. Close the stock legs before expiry, or know what a physically settled option does.
Ten names hold a bit over half the index, so a ten leg basket is a partial copy
What the Indian market adds
- Settlement. Nifty options are European and settle in cash. Stock options settle by physical delivery, a rule SEBI phased in for all stocks by Oct 2019. A stock straddle that is held to expiry finishes with one leg in the money, except exactly at the strike, and that leg means delivery.
- Tenor. Stock options expire monthly. The index has weeklies, but a dispersion pair needs one expiry for both sides. Bank Nifty has monthly expiries only and far fewer stocks than the Nifty, which makes a small basket cover more of the index. This lesson has no weights file for it, so it does not run that case.
- Lots. Each stock has its own lot. The vega match is rounded to whole lots.
- Bans. A stock near its market wide position limit goes into a ban period, and new positions in it are not allowed. A rebalance can be impossible on the day it is needed.
- Costs. STT on options is 0.15% of premium on the sale side from 1 Apr 2026, and there is the exchange charge on every leg. A ten name basket has twenty two option legs against two for the index alone.
Every name adds two legs, and each leg has its own spread and its own charge
Worked example: the 30 Jun 2026 expiry, on paper
On 10 Jun 2026 the index leg was priced at 15.03%. The fifty stock volatilities, weighted, averaged well above that. By the expiry, thirteen sessions later, the index had realised well under what it was priced at, while the weighted stock volatilities had realised about what they were priced at. Here is the paper result in volatility points per unit of vega, before delta hedging, costs and rounding.
The index leg did the work: the stock legs were roughly flat
Read 2 more paragraphs, 177 words
Three cautions on how to read that bar chart. The windows are thirteen close to close returns, a short stretch, and the realised volatility of a name over so few days is noisy. The weights are held at their 10 Jun values although they drift. And names with a one day jump above a quarter, which is a corporate action and not a market move, are left out of the window and the rest are renormalised. The five names flagged in this data set are the reason for that rule. None of this changes the arithmetic. It changes how much weight a single fortnight can carry.
Almost all of it came from the index leg. The stocks moved about as much as the options said they would. The index moved less than they said, because the stocks did not move together. The realised correlation over those thirteen sessions was well below the implied one. One expiry on a thirteen session window is a single draw, so the number shows how the arithmetic works, not what to expect.
What would be watched while it is open
Three things move day to day. The implied correlation changes with the index volatility and the stock volatilities, as the sensitivity chart of the last lesson showed. The delta of each leg changes with the index and the stock, and with time, which is where the vanna and charm of module E05 come in. And the margin changes with the short index leg, which is where a rise in index volatility hurts before any loss is booked. A record of the position that shows only the profit of the day hides all three.
Read one more paragraph, 42 words
Listed variance swaps do not trade in India, so a basket of index and stock options is the public way to hold this view. That is also why the legs, the lots and the costs matter so much here: nothing replaces them.
The profit by correlation
Keep the stock volatilities at their implied levels and let the realised correlation take other values. The index volatility that follows is the square root of B plus rho times A minus B. The profit is the implied index volatility less that.
The trade breaks even at the implied correlation, gains below it and loses above it
Read one more paragraph, 72 words
The bar for the realised correlation of the example, the thirteen sessions to expiry, gives nearly the paper profit, as it should. Now read down the list. The result is zero at the implied level, positive below it and negative above it, and it bends: the same distance of correlation is worth a little more on the gain side. The loss side is still the one that arrives with a falling market.
When correlation spikes
Correlation tends to rise when markets fall, and a sharp fall lifts index volatility with it. The stock volatilities rise too, which cushions the long legs, but the part that is not cushioned is the rise in correlation. Realised correlation was 0.31 on 6 Mar 2026 and 0.60 on 8 Apr 2026. The series below shows the Nifty's twenty session realised volatility over the same weeks, which reached 29.46 per cent on 8 Apr.
Index volatility spiked in April 2026, and with it correlation
Appendix: what I would check before sizing one
First, whether the correlation I am short is high or low against its own history, as in the last lesson, and whether the history I use has enough expiries. Second, the loss at a realised correlation well above the implied one, with the index volatility that goes with it, set against the capital the position ties up. Third, each stock's calendar: results, corporate actions and ban lists inside the expiry. Fourth, the cost of every leg, including the exit. Fifth, a plan for a physically settled leg that finishes in the money. None of this makes the trade right or wrong. It sizes the ways the paper result of a quiet fortnight can differ from a stormy one.
Read one more paragraph, 64 words
A position sold at the June levels and facing that stretch would have been on the loss side of the table above, with the index realising far above what it sold. The implied levels of March and April are not in this data set, so this is a what if, not a record. It is the same trade as June, with the other draw.
The common mistake
The claim: Dispersion is a market neutral trade, so it makes money whatever the market does.
It is neutral to the direction of the index at the start, because each leg is delta hedged. It is not neutral to correlation, and it is short it. On 8 Apr 2026 the twenty session realised correlation reached 0.60 while the index realised 29.46 per cent. At that reading the paper result by correlation is a loss. The June paper profit and the April what if loss are the same trade with different draws. Ask how the trade behaves when everything moves at once.
Key takeaways
- Dispersion sells index volatility and buys stock volatility, matched in vega.
- After delta hedging, the main exposure left is to correlation.
- Profit is zero at the implied correlation, positive below it, negative above.
- On 10 Jun 2026 the paper result to expiry came from the index leg.
- Correlation spiked in April 2026, the draw that hurts a short correlation trade.
- Physically settled stock options need a plan for legs that end in the money.
Before you trade. Derivatives trading carries a high risk of loss. SEBI's study "Profitability of Individual Traders in the Equity Derivatives Segment (FY25 to FY26)" (August 2026) found that 87.7% of individual traders in equity F&O made net losses in FY26, with aggregate net losses of Rs 91,685 crore. Read the SEBI study