Guide · Margins and risk · Updated 2026-09-19

The spread benefit is a lower risk estimate, not a discount, and it is withdrawn before your position changes

Most writing on this subject describes the benefit as something the system gives you and then takes away, which makes it sound negotiable and makes the expiry day rise sound punitive. Neither is true, and the belief is why the rise surprises people who have read about it. The benefit is an output, and understanding what produces it is the only way to know in advance when it stops being produced. That turns out to be a date you can write in a calendar rather than an event you discover from a message on a Tuesday morning.

A spread to you and a spread to the risk model are different objects

The word spread carries a lot of meaning in a trading conversation and almost none in a margin file. A trader means two positions held together for a combined payoff. A risk model has no access to intent. It receives a portfolio at the end of a session and recognises an offset only where its own construction finds one.

Four conditions narrow the set, and each of them is narrower than the everyday usage.

One underlying. The offset is found within a single instrument family. A long index position against a short position in a correlated index is not a calendar spread and is not handled here at all. It belongs to the separate cross margin framework, which the October 2024 circular notes is fully revoked at the start of the first of the expiring correlated indices, on the same reasoning.

One portfolio. The legs have to sit in the same client account at the same clearing member. Two accounts holding one leg each are two portfolios with no relationship, however deliberate the pairing looked when it was put on.

Matched on delta, not on lots. This is where most of the surprise lives. The model matches the delta the portfolio carries in each expiry month, so one lot against one lot is rarely a whole spread. In the worked example below, the near leg carries a delta of 0.377 and the far leg 0.497, so only 0.377 of the position is matched and the rest of the far leg stands alone.

Recomputed every cycle. The pairing is not stamped on the position at entry. It is rebuilt each time margins are computed, from whatever the portfolio then contains. A spread order placed at the exchange is an execution convenience and confers no status. The margin treatment comes from the resulting portfolio, and a change in the far leg's delta can shrink the match without anything being traded.

The benefit is an output of a scenario grid, not a percentage anybody grants

The mechanism is a scenario revaluation. The model takes the underlying and moves it in steps up to a published price scan range, and at each of those price points it moves implied volatility up and down by a volatility scan range. Two further scenarios push the price twice the scan range in each direction and count a fraction of the resulting loss, on the view that a move that large is real but should not dominate. The portfolio is revalued in every one of those states, and the largest loss across the set is the worst scenario loss.

Run that grid three times on the same eight day old position: once on the short near leg alone, once on the long far leg alone, and once on the pair together.

The short leg's worst case is a sharp rise. The long leg's worst case is a fall with volatility coming in. Those are different columns of the same grid, and a portfolio can only be charged for one scenario at a time. That single fact is the whole of the benefit: 1,05,156 on these inputs, or 82.2 per cent of what the legs would cost separately.

Nothing was granted. There is no rate in a circular that produces 82.2 per cent, no schedule to look it up in and nobody to appeal to when it changes. The number is a difference between two revaluations of the same portfolio, and it moves when either revaluation moves.

Two components are then added on top. A calendar spread charge, because the near and far contracts do not in fact move identically and the grid, which shifts one underlying price, does not capture that. And extreme loss margin, which sits outside the scenario grid entirely and is levied at a published rate on the short position rather than computed from any revaluation.

What the circular changed, and the sentence that does the work

The October 2024 circular states its reasoning before it states its rule. Clause 5.2.1 observes that expiry day carries significant basis risk, where the value of a contract expiring on the day can move very differently from the value of similar contracts expiring in future. That is the economic claim, and it is correct: a contract with hours left and a contract with a month left stop being substitutes.

The rule that follows is narrow and mechanical. Clause 5.2.2 directs that on the day of expiry the worst scenario loss shall be calculated separately for contracts expiring on the given day and for the rest of the contracts. It then draws the two consequences out explicitly: because the worst scenario loss is computed separately, an additional calendar spread margin is not applicable to contracts expiring that day, and where one leg of a futures calendar spread expires on the day, the extreme loss margin is computed without treating those futures as an offsetting spread position.

Read that as an instruction to the model rather than as a change to a rate, because that is what it is. No percentage was altered. The portfolio was cut in two before the grid was run. Once the two halves are valued separately, the loss on a rise and the loss on a fall are both charged, and the requirement reflects a market that goes up and down on the same day.

Clause 5.2.3 is the part almost nobody quotes and the part a trader can actually use. It states that spreads involving contracts other than the expiring one keep their treatment, and illustrates it with three monthly expiries falling on the 29th, 30th and 31st. Spreads pairing the 29th with the 30th, or the 29th with the 31st, lose their treatment on the 29th. A spread pairing the 30th with the 31st keeps it on the 29th. The withdrawal is leg by leg, not portfolio wide, and it is triggered only by holding the contract that expires that day.

The same position, computed twice

Run the grid again on the morning the near leg expires, with the portfolio partitioned as the circular directs. Nothing has been traded. The far leg is four weeks younger and the near leg has hours left.

Two things drive the jump and they are worth separating. The partition itself contributes most of it: the expiring leg's worst case of 1,02,058 and the far leg's worst case of 27,655 are added, when the pair valued together on the same morning would have produced 27,626. The rest is the additional extreme loss margin of 2 per cent that clause 5.6.1 applies to short options expiring on the day, which on this notional is 39,000.

That 2 per cent is widely misreported as 2 per cent of the margin. It is 2 per cent of the notional value of the short option position, and on a contract of this size the difference between those two readings is more than a lakh of rupees. It applies both to short positions open at the start of the day and to short positions created during the day in contracts expiring that day.

Had calendar spread treatment survived into expiry day, as it did before February 2025, the same portfolio on the same morning would have carried about 1,05,810. So 1,01,903 of the 1,40,381 rise comes from the partition alone, and the remainder from the expiring leg sharpening as it runs out of time.

Single stock derivatives joined the rule in May 2026

For the first fifteen months the measure applied to equity index derivatives and to nothing else. Clause 5.2.4 of the October 2024 circular says so in one line, and clearing corporations implemented it from 10 February 2025 against the circular's stated date of 1 February 2025.

That changed. A SEBI circular dated 5 February 2026, carrying reference HO/47/15/11(2)2025-MRD-TPD1/I/4226/2026 and titled as a review of calendar spread margin benefit in single stock derivatives on expiry day, extended the same treatment to single stock futures and options, with effect from early May 2026. Exchange and broker notices published at the time gave the first applicable date as 4 May or 5 May 2026, so confirm the exact first expiry against the current exchange notice rather than assuming.

This is the fact that dates most competing material. A page written at any point between February 2025 and February 2026 will tell a reader, correctly for its time, that stock spreads keep their benefit through expiry while index spreads do not. That distinction no longer exists, and a trader working from it will size a stock calendar spread against a requirement that has not applied since the spring.

The single stock case is also harsher than the index case for reasons that have nothing to do with margin. Single stock derivatives settle by delivery, so the expiring leg does not vanish into a cash difference; it becomes an obligation in the underlying shares, with its own margins in the days before expiry. And the underlying can enter a position limit ban period, during which fresh positions are restricted, which is precisely the window in which a trader who has left the roll late needs to open a new far month leg.

The erosion before the cliff

The rule is a step, but the offset was already shrinking before it landed, for an entirely different reason. As the near leg loses time it becomes sharper: its value responds more violently to a move through the strike, while the far leg, with weeks left, does not. The two legs stop matching each other's behaviour gradually, and the model sees that in the grid long before any partition is applied.

From 25 days out to one day out the offset falls from 86.5 per cent to 79.0 per cent, and the total requirement barely moves, from 63,856 to 68,234. A trader watching the requirement rather than the mechanism sees nothing happening for three weeks, and then sees the whole move in one file.

That is the shape to hold in mind. The slope is risk, and it is legible in the numbers if you look. The step is a rule, and it arrives at full size whatever the market has done, on a date fixed months in advance.

The call arrives on a position that has not moved

The timing catches more people than the arithmetic. Margin requirements for a session are carried in the file the clearing corporation cuts after the previous session closes. The requirement that applies on expiry day morning therefore exists from the evening before, which is when many Indian brokers ask for the money. A trader who has read that the benefit goes on expiry day, and who plans to deal with it on expiry day, finds the demand already sitting in the account the night before.

The position, meanwhile, has done nothing. It has not moved against the trader, it has not been adjusted, and on the trader's own reading of its risk it is no more dangerous than it was a week ago. The reading is not unreasonable. It is simply not the reading the risk model is making, and the model's reading is the one that has to be funded.

The exit is also worse than it looks. Closing the expiring leg leaves an outright position in the far leg, which is a different trade at a different risk. Closing the far leg leaves an uncovered short running into its own expiry, which is considerably worse than the margin problem. Preserving the risk profile means closing both, and on that morning one of the two is a contract with hours to live while the other is a far month contract whose book is often the thinner of the two. Forced unwinding of the illiquid leg is the cost that does not appear in any margin table.

Falling short has its own consequences. A margin shortfall is reported and penalised under the existing framework, and the penalty attaches to the shortfall rather than to the reasoning behind it.

Calendar spreads and vertical spreads are bisected differently

The partition is made on the expiry date and on nothing else, which decides in advance which structures survive it.

A vertical spread has both legs in one expiry, so both land in the bucket for contracts expiring that day and the model still revalues them together. In the computation, the short leg alone carries 1,02,058 on expiry morning, and the two legs together carry 14,970, which is 85.3 per cent less and sits just under the 15,000 that is the structure's true maximum loss, being the width of the strikes multiplied by the lot. The model is charging what the position can actually lose.

The requirement still rises, from 49,502 to 92,970, which is 1.88 times. But almost all of that is the additional extreme loss margin on the short leg rather than a lost offset. The structure is intact; the rate applied to it went up.

None of this makes a vertical spread benign into expiry. Both legs are settling on a price that is being decided in the session, a strike between them is where the outcome flips, and in single stock contracts one leg can be exercised into delivery while the other expires worthless. Those are different problems from the margin problem, and they do not show up in a scan range.

The benefit moves with volatility, not only with time

The offset is produced by a risk model whose inputs the clearing corporation revises, and that has a consequence most descriptions of the benefit omit. The price scan range widens when realised volatility rises. Implied volatility changes the shape of every revaluation. Deltas move, and the matched portion of the position moves with them. None of this requires a trade.

Read the last two columns against each other. The share of risk offset rises from 80.0 per cent to 84.8 per cent as volatility climbs, while the rupee requirement for the pair rises from 18,989 to 38,071, which is slightly more than double. The percentage improves and the cash requirement roughly doubles, at the same time, on a position nobody touched.

That is why a published percentage would be useless even if one existed. Any sentence of the form the benefit is about eighty per cent is compatible with the requirement doubling. The only figure that means anything is the rupee requirement under the current parameters, and the only safe assumption about it is that it will be larger in a fright than it is in a calm.

What a trader can actually do about it

Roll before the file is cut, not on the morning. A spread between two contracts that are not expiring that day keeps its treatment, which the circular states directly in its own illustration. Moving the position one expiry out ahead of the cut off produces a requirement of 62,344 in this example against 2,07,713 for holding through, a difference of 1,45,369. The roll has a cost in spread and slippage. It is smaller than the difference, and it is paid at a time of your choosing.

Size against the requirement that applies after the benefit is gone. On ten lakh of usable margin, the benefited requirement supports 14.9 lots and the expiry day requirement supports 4.8. A trader who sizes to the first number is carrying about three times the position the worst morning can fund. The funded figure has to be the one that exists on the worst morning, not the best.

Write the date down when the position is opened, not when it is questioned. The relevant date is not expiry day. It is the close of the session before it, because that is when the file carrying the expiry day requirement is produced. For a position held across several expiries there is one such date per cycle, and they are all known in advance.

Check the far leg's book before you need it. The exit that preserves the position's shape requires trading both legs, and the far month contract is usually the less liquid of the two. A structure that can only be unwound in a thin book is carrying a cost that no margin file shows, and the current scan ranges are published, so a requirement computed from last quarter's parameters is a guess and in a volatile week a low one.

Where this goes wrong in practice

Treating the benefit as permanent because it has been there for weeks. The erosion table shows a requirement that barely moves for three weeks and then moves entirely in one step. Stability up to the step is not evidence about the step.

Assuming a two leg order buys a margin status. It does not. The model rebuilds the match from the portfolio at each computation, on delta, and the trader's intent is not an input at any stage.

Reading the additional 2 per cent as 2 per cent of the margin. It is 2 per cent of the notional value of the short option position expiring that day, which on an index contract is a much larger number than the misreading suggests.

Working from a page written before February 2026 on single stock spreads. The treatment that separated index from stock derivatives ended in the spring of 2026. Material that still draws that line is describing a position that has lapsed.

Planning the roll for expiry day morning. By the morning the requirement is already in the account, the expiring leg has hours left, and the far leg has to be traded in whatever book exists. The decision was available for weeks and was taken on the one day it could not be taken cleanly.

Confusing a spread with a hedge. A position that reduces your economic risk and a position the model recognises as an offset are different sets, and only the second reduces what you have to fund. Everything outside the intersection is a risk view you are carrying at full margin.

What the benefit is actually for

Margin exists to cover the loss a clearing corporation would have to absorb if a member failed. The offset on a spread is not a courtesy extended to hedged traders; it is the honest answer to that question when the portfolio genuinely cannot lose both ways at once. When the two legs stop being substitutes, as they do in the last hours of one of them, the honest answer changes, and the framework changes with it.

Which means the number to plan around was never the benefited one. It was always the requirement the position carries on its worst scheduled day, and that day is printed on the contract. Working from it is not conservatism. It is the only reading of the position that is available on every day it exists, rather than on most of them.