Guide · Regulation and market structure · Updated 2026-09-23

The retail derivatives rules were written from the regulator's own evidence, and they changed the cost of trading, not its arithmetic

Most coverage tells this as one crackdown provoked by one headline, the 93 per cent. The documents describe a sequence with a loop in it. The first loss study, in January 2023, produced a risk disclosure and a duty on the largest brokers to keep every client's profit and loss in a fixed format, and the later studies report those same fields. The October 2024 package was drafted from expiry day trading evidence in a July 2024 consultation paper and issued eight days after the second loss study. The 2025 rules changed the expiry calendar and the unit in which open interest is measured. The 2026 changes raised the transaction tax again and moved the closing price that expiring contracts settle against.

This page is not another summary of the losses; those are in the loss report explained and the base rate and its mechanisms. It is the policy sequence: each measure, its circular and effective date, the mechanism in the regulator's own evidence it was aimed at, and what the data shows after it.

The loss figure most pages quote was restated by the regulator

In July 2025 the regulator published a short update putting FY25 net losses of individual traders at ₹1,05,603 crore, 41 per cent above FY24's ₹74,812 crore, with 91.0 per cent of traders loss making. It used the top 13 brokers, whose clients numbered about 96 lakh of roughly 107 lakh traders in the market. The August 2026 study returned to the 15 brokers used in September 2024 and restated FY25 at ₹1,11,788 crore, 5.9 per cent higher, with a footnote giving the larger broker sample as the reason. It also restates the FY25 loss share as 90.9 per cent on 98.10 lakh traders. Two participation counts now circulate, and both are the regulator's: its press release gives the broker sample, 98.1 lakh falling to 78.6 lakh, about 20 per cent, while the study's summary gives every trader on both exchanges, 106.2 lakh falling to 87.5 lakh, 18 per cent.

The restatement decides the headline. Against the restated base, FY26 losses fell 18.0 per cent. Against the figure still printed on most pages, the same FY26 number is a fall of 13.2 per cent, a comparison across two samples that measures nothing. The study adds two cautions of its own. Its sample covers about 90 per cent of individual traders, and if the rest behave alike the population loss is about 1.11 times the reported one. And a lower loss share in a year when a fifth of traders left, it says, need not mean that outcomes improved.

Four studies, and the record keeping rule behind the last three

Three things in that table are easy to miss. The samples differ, ten brokers, then fifteen, thirteen and fifteen again, so rupee totals compare cleanly only inside one study while loss shares barely move between samples. "Individual" includes HUFs and NRIs throughout, and the 2026 study also counts sole proprietorships and individuals trading through portfolio managers, so it is not a pure retail count. And every figure is realised: positions still open at the year end are left out.

The fourth is structural. The circular of 19 May 2023 that put the first study's findings on every login screen from 1 July 2023 also required qualified stock brokers to keep each client's realised profit and loss, split into brokerage, exchange fees, stamp duty, the regulator's fee, STT and GST, with age, gender, income band and PIN code, for at least five years. Those are the fields the 2024 to 2026 studies report. The disclosure was the visible measure. The record keeping duty is the one that made measurement routine.

The October 2024 package was drafted from expiry day evidence, not from the loss total

The circular of 1 October 2024, SEBI/HO/MRD/TPD-1/P/CIR/2024/132, does not cite the loss study published eight days earlier. It rests on an expert working group, the regulator's secondary market advisory committee and a consultation paper of 30 July 2024, and that paper's evidence is about expiry day. It found three minute windows before the close in which traded volume ran as high as 822 times the change in open interest, reported the exchanges' estimate that a retail position was held for about 30 minutes, and showed the time value in the nearest in the money call and put on the benchmark index falling from ₹126 a unit the day before expiry to ₹5.80 at 3 pm and ₹1.20 at 3.15 pm on the day. Its loss evidence was the January 2023 study and one exchange figure: 92.50 lakh individuals traded index derivatives on the larger exchange in FY24 and lost ₹51,689 crore before costs, about 85 in every 100 of them at a net trading loss.

The circular sorts its own six measures by purpose: four aimed at expiry day activity, one at keeping the product suitable for the people trading it, one at basic risk hygiene. The consultation had proposed seven, and the gap between draft and rule shows what the regulator was prepared to trade away.

Two of the softenings bear directly on retail cost. The contract floor stopped at ₹15 lakh instead of stepping up to ₹20 lakh, and the expiry margin went from three extra points the day before expiry and eight on the day to two points on the day alone, charged only on short options. The measure dropped outright, strike rationalisation, was aimed at the far out of the money strikes the paper found drawing fresh positions on expiry day.

Measure by measure: the date, the rule and the mechanism it was aimed at

Grouped by what they act on, the rules pull six levers. The ticket: a contract worth at least ₹15 lakh, three times the old floor, which the circular itself calls an inbuilt suitability test. The calendar: one weekly contract per exchange, then every expiry of an exchange on one weekday, then only Tuesday or Thursday. The first session under the November measures was 21 November 2024, because 20 November was a market holiday. The collateral: premium paid before the trade, no spread offset for a leg that dies that day, two extra points of margin on short options on expiry day. How the offset is computed and why its removal raises the requirement without the position changing is in the spread margin guide.

The ruler: since 2025 the regulator measures open interest in delta adjusted futures equivalents, so a far out of the money option counts for a fraction of one at the money, and its index option limits are rupees of delta rather than rupees of notional. The September 2025 circular is candid about why the intraday version arrived: it cites observed instances of outsized intraday positions built by certain entities in index options on expiry day. The toll: STT raised in October 2024 and again in April 2026. The price: since 3 August 2026 a closing auction sets the close of stocks with derivatives, and with it the prices expiring stock and index contracts settle against. The regulator is now consulting on whether settlement should blend the auction with the last half hour of continuous trading, or set it aside for at least a year.

What the regulator measured afterwards: fewer traders, not less trading

The pattern is consistent, and the regulator reads it the same way. The access measures removed people at the margin they were aimed at: traders with annual turnover below ₹10,000 fell 37 per cent in FY26, while those turning over ₹1 crore to ₹10 crore rose about 1 per cent. Those who stayed traded more. Premium turnover in index options fell and then recovered past its starting level, reaching ₹1,18,474 crore a day in March 2026. Expiry day concentration eased, but 97 per cent of index option turnover still sat within a week of expiry. On 18 September 2026, a Friday two sessions before the weekly expiry, 91.7 per cent of the index option contracts traded on the larger exchange were in the contract expiring on 22 September, measured from that day's derivatives bhavcopy. The study's own conclusion is that the measures moderated participation without fundamentally changing behaviour.

Two recomputations are worth recording, because they are the kind that catch errors. The study prints the first fall in daily premium as 17.4 per cent; its own rupee figures, ₹65,935 crore and ₹54,771 crore, give 16.9. It prints the fall in index option traders from 63.70 lakh to 46.31 lakh as 17.09 lakh and 26.8 per cent; the two counts differ by 17.39 lakh, 27.3 per cent. Neither changes the reading, and both are reasons to work from a document's figures rather than its summary.

Measured: the client share of the open book dipped and came back

The studies count people and rupees. The exchange publishes a third quantity every day: open interest in index options held by four categories, client, domestic institution, foreign portfolio investor and proprietary, on the long and the short side. 2,642 of those files, 1 January 2016 to 18 September 2026, give the client category's share of each side of the book.

The rise came first and the turn came before any rule. The client share of long index option positions averaged 42.6 per cent in 2016 and peaked at 65.9 per cent in July to September 2022; the short side peaked at 69.5 per cent in July to September 2021. Both were well off their peaks when the October 2024 circular was issued, which matches the regulator's own remark that participation was already moderating before the measures.

In the initial contraction the client share fell by 4.0 points on the long side and 4.2 on the short, while the proprietary share rose by 3.3 and 3.0. Since April 2026 the client share has averaged 58.8 and 64.7 per cent, within 2.2 and 0.9 points of where it started, while the number of individuals trading fell 18 per cent between FY25 and FY26. The leavers were overwhelmingly small and occasional traders, and a share that barely moved is consistent with their having held little of the overnight book.

Timing can be tested more finely. For each date a measure took effect, compare the mean share over the 20 sessions from that date with the 20 sessions before it, and ask how unusual the change is against the same statistic for every session from January 2023 to July 2026, 870 of them, measure or not. An ordinary 20 session change is 0.83 points on the long side at the median and 2.10 at the 90th percentile; on the short side, 0.90 and 2.98.

Two of the 22 changes clear the 90th percentile, against 2.2 expected by chance. No single measure moved the client share of the open book by more than an ordinary month does. The shift shows only over quarters. Windows of 60 sessions give falls of 2.9 and 4.7 points around 21 November 2024, but windows that long run into the next two measures and cannot be pinned on one.

An open interest share is not a head count, and a contract is not a unit of risk

The share series is a real measurement of a real quantity, and it is easy to read it as something it is not. Four gaps separate it from the studies.

It is a snapshot at the close. Open interest is what remains open after the session. The consultation paper's 30 minute holding period and the study's finding that 59 per cent of FY26 index option turnover was in contracts on their last day both describe trades that leave no trace in an end of day file. The traders the rules were aimed at are the ones the file sees least.

The client row is not the retail row. The file has four rows. The regulator's study reports corporates, trusts and partnership firms as categories of their own, and in a four row file they can only sit under client, along with wealthy individuals. A share of that row is not a share of the individuals the studies count.

It is one exchange. The files come from the larger exchange's clearing corporation. The other exchange's index options grew after November 2024 while the larger exchange's shrank; its premium traded on non expiry days rose from ₹1,101 crore a day to ₹6,917 crore, 6.3 times, in the first four months. None of that book is in these files.

A contract is not a quantity of risk. The median open index option position on the long side was 21.6 million contracts from September to 19 November 2024 and 8.2 million in February and March 2025, 62 per cent lower, over the weeks in which every index contract was recut from a value of ₹5 to 10 lakh to at least ₹15 lakh and the extra weekly series were withdrawn. A larger contract needs fewer of them for the same exposure, so the count says little about positions. A share survives this, because every row of a day is counted in the same contracts; a count across a lot change does not. The regulator reached the same conclusion from the other side when it moved to delta for its own limits, while the participant file still weights every contract equally. The delta measure is worked through in the ban period guide and the position limits guide, and the lot arithmetic in why lot sizes are revised.

What regulation can change, and the arithmetic it cannot

Every option has a buyer and a writer and every future a long and a short, so before costs one side's gain is the other's loss. The FY26 study says so, and its own figures show it to the crore. Its ledger of gross realised results for FY25 and FY26 combined, drawn from exchange data on every participant, has individuals losing ₹1,70,125 crore while proprietary traders, foreign investors, corporates, mutual funds, partnership firms, others and domestic institutions gained ₹1,70,124 crore between them.

That identity is what no circular reaches. A rule can change who may enter and at what ticket, how often an expiry arrives, how much collateral a position needs and what the exchange and the government charge per trade. It cannot change the sign of the transfer between categories, which is set by who trades against whom, at what speed and with what information. The study reports that 99 per cent of the gross profit of foreign investors and proprietary traders was earned by entities placing algorithmic orders, and that the ten largest proprietary traders took about three quarters of that category's profit.

Costs are the part a rule can move, and the rules moved them up. In FY26 the individuals in the regulator's broker sample lost ₹66,896 crore on their trades and paid ₹24,789 crore in transaction costs, which together make the ₹91,685 crore net loss; costs were 27 per cent of it. STT paid by individuals rose 35 per cent in FY26 while turnover fell, and the April 2026 increase falls on the next study year in full. A higher toll lowers the expected result of every trade in proportion to what is traded. It deters the most frequent traders most, and it improves no remaining trader's arithmetic. The cost stack line by line is in the real cost of an Indian trade.

Reading the next rule

The sequence gives a way to read whatever comes next. Ask which lever a rule pulls, the ticket, the calendar, the collateral, the ruler, the toll or the price, and which mechanism in the regulator's evidence it names. Then ask which quantity would show whether it worked, because the studies count heads and rupees and the exchange files count open positions, and a rule can move one without the other. Two items are open as this page is written. The consultation paper of 12 September 2026 proposes how expiring contracts should settle now that an auction sets the close, with comments due by 3 October 2026. A paper of 4 December 2025 proposed moving trading members' index option limits onto the same delta basis as clients'; no circular implementing it appears in the regulator's derivatives listings as of this date.

For a trader the practical content is narrower. The studies describe a population, and the rules change the cost of belonging to it. Neither says whether a particular trade has an edge after costs, which is the only question that decides an individual result. Answering it is a matter of method and of measuring your own record honestly, which is where our curriculum spends its time.