Guide · Derivatives mechanics · Updated 2026-09-19

A futures price above spot is an arithmetic statement about funding, not a view on direction

The most repeated error in retail derivatives material is not about a formula. It is about what kind of object a futures price is. Read a dozen pages and you will find it described as what the market thinks the index will be worth at expiry. It is not, and the difference changes what you conclude every time the number moves.

A bound, not a prediction

Consider two ways to own the index at a fixed date. Borrow money, buy the index at spot, hold it, collect whatever it distributes, and repay the borrowing at that date. Or buy the future and set the cash aside. Both end holding the index and nothing else. Two positions with identical terminal outcomes and no residual risk cannot cost different amounts today.

That equality is the whole mechanism. The forward price equals spot grown at the funding rate over the period, reduced by the distributions the spot holder receives and the futures holder does not. No expectation appears in the derivation, because none is needed. It runs on prices that exist now and cashflows that are contractually determined.

It is a bound rather than an identity because the enforcing trade is not costless. Each side has to be executed, financed and margined, and those costs widen the relation into a band. Inside the band there is no trade to do, so the price can sit anywhere. That band is where the subject actually lives, and most generic pages stop before reaching it.

A futures price above spot is a statement about funding

When funding costs more than the asset distributes, the forward sits above spot. That is the ordinary condition for an equity index in India, where the cost of money comfortably exceeds an index dividend yield that ran at a median of 1.28 per cent across this sample.

Now suppose every participant became convinced the index would fall five per cent before expiry. Nothing in the derivation changes, because it never used a view. The spot price would fall to reflect the conviction and the forward would fall with it, still above the new spot by the same carry. Expectation acts on spot. Given spot, the forward is arithmetic.

The corollary is the part worth keeping. If a futures price moved to reflect a directional view independently of spot, it would leave the band, and the trade that closes the gap would appear. Direction cannot get into the basis without creating the incentive to remove it. That is not a claim about how sensible participants are. It is a claim about what survives.

What the basis contains, itemised

The basis is the futures price minus the spot price of the same underlying at the same instant. Each component has a different source, a different sign, and a very different degree of observability.

Only the first two are in the textbook relation. The third is why measured bases do not sit on the textbook value, and it carries almost all the India specific content.

Work one measured session through. On 2026-09-03 the near month contract closed at 24,000.10 with the index at 23,873.45, a basis of +126.65 points or 0.5305 per cent, with 26 calendar days to the 2026-09-29 expiry. The exchange's own index file put the trailing yield at 1.19 per cent.

Run the arithmetic backwards and the session implies an annualised carry of 7.43 per cent. Add the published trailing yield, assuming distributions accrue evenly through the year, and the implied funding rate is 8.62 per cent. Both are upper estimates, because Indian distributions do not accrue evenly and a window in which few companies go ex carries less than a proportional share of the annual yield.

The carry period is an exchange calendar, not a month

Every annualised figure above divides by a number of days, and in India that number is set by a rule that changed recently. Monthly derivatives contracts on this exchange expired on the last Thursday until the standardisation effective from 1 September 2025, and on the last Tuesday afterwards. The regulator required each exchange to pick one of two fixed weekdays and stay on it.

That is checkable against the exchange's own files rather than against anybody's description of the rule. Every one of the 13 monthly expiries before the cut fell on a Thursday, and 12 of the 13 after it fell on a Tuesday. The remaining one fell on a Monday, 2026-03-30, because the Tuesday was not a trading day.

The carry period therefore takes the values 28, 29, 33, 34, 35 days in this sample and never equals a calendar month. The cycle that straddled the change ran 33 days, because it began on a Thursday and ended on a Tuesday five weeks later. Any page that annualises by dividing by thirty gets a different answer from the one the contract specifies.

The spread across that table is 2.53 percentage points of annualised carry, produced entirely by the choice of divisor. A reader comparing an implied funding rate against a money market rate can be wrong by more than the thing they are trying to detect, purely by taking the period from a calendar instead of the contract.

What an ex date does to the basis, and why nothing happened

The holder of the index receives its distributions. The holder of a future on it does not. So when a constituent goes ex, the spot index falls by roughly the amount that went ex and the futures price does not follow, because the future never had that entitlement priced into it. The basis widens by approximately the distribution, on the morning of the ex date, mechanically.

Nothing was learned and nobody revised a view. Watching the basis widen that morning and calling it a shift in positioning mistakes an accounting step in one leg for information. The signature of the real thing is that it concentrates in the months when Indian companies actually go ex, the stretch after the March year end when final dividends are approved.

Grouping April through July against October through January, the heavy months average 4.34 per cent of annualised carry against 7.80 per cent, a gap of 3.46 points. Two years is a short sample and funding has its own seasonality, so this is consistent with the distribution leg rather than proof of it. What does not depend on sample size is the direction: a lower basis in the dividend heavy months is what the mechanism predicts, and reading it as seasonal pessimism would be reading a corporate action calendar as a mood.

The bound has two sides and only one of them is easy to enforce

To pull an expensive future down you buy the index and sell the future, hold both to expiry and take the difference. That needs capital, margin on the short leg and a financing rate. It is demanding but available to anyone with a balance sheet.

To pull a cheap future up you do the reverse: sell the index you do not own and buy the future. That needs the constituents borrowed first. In India the securities lending and borrowing route is narrow and thinly used, and cash market short selling outside it is intraday for most participants, which does not span a carry period. The lower bound is enforced by a trade that frequently cannot be done at all.

The asymmetry is visible in the measurement. Across 446 sessions more than four days from expiry the basis was negative on 24, or 5.4 per cent. Annualised carry sat between 4.26 and 8.05 per cent across the middle half of sessions, with a fifth percentile of -0.12 and a ninety fifth of 11.79. That is not a distribution centred on a funding rate with symmetric noise around it. It has a floor enforced weakly and a ceiling enforced hard.

Those five rows are the honest content of the residual. A page that presents the carry relation as an equality and then expresses surprise that the market disobeys it has omitted the row that explains the difference. Position limits matter most, because they remove the enforcing participant at exactly the moment the gap is widest, and the mechanics are covered in the guide to market wide position limits and the ban period.

Convergence is the only certain part of the subject

Everything above is a band, a residual or an estimate. One thing is not. At expiry the contract settles against the closing level of the underlying, so the basis is zero by rule rather than by tendency. That terminal condition is written into the contract, which makes the basis a decaying quantity with a known endpoint.

The mechanism is explicit in the exchange's own file. The published futures close equalled the daily settlement price on 503 of 528 sessions, which is every non expiry session in the sample. On each of the 25 expiry sessions the settlement price of the expiring contract equalled the closing index level exactly, to the paisa. What remains on the final session is the gap between the last futures print and that close, averaging 8.11 points with a standard deviation of 11.71, against 111.2 points when a full period remained.

Two consequences follow. Anything inferred from a basis must be expressed per unit of remaining time, because a basis in points is not comparable across days to expiry. And a basis that fails to converge is not a market judgement, it is a data problem, usually a stale print or a mismatched timestamp between the legs.

Three indices, one session, three different carries

The cleanest demonstration that the basis is not a mood reading is to hold constant everything a mood would depend on. Take one session, one expiry date, one currency and one funding market, and read the carry on three different indices.

The spread across the three is 10.65 percentage points of annualised carry on a single afternoon. A directional reading would have to claim the market was simultaneously enthusiastic about one index, moderate about another and negative about a third, at the same instant, on overlapping constituents. The funding reading needs no such claim: the three differ in what they distribute, in how cheap the cash basket is to assemble, and in how easily each can be borrowed and shorted.

Read the last column carefully, because it is where the residual becomes visible. An implied funding rate of 0.24% is not a rate anybody in India can borrow at. It is what the arithmetic returns when the lower bound is not being enforced, because the trade that would enforce it cannot be assembled in the constituents of that index at acceptable cost. The number is not telling you what money costs. It is telling you the arbitrage is absent.

The spot reference changed on 3 August 2026

A basis is only as well defined as the two prices it subtracts. In India the cash leg's definition changed six weeks before this page was written, and most material on cost of carry has not caught up.

From 3 August 2026 the official closing price for stocks in the derivatives eligible universe is the single equilibrium price of a closing call auction held after continuous trading ends, replacing the volume weighted average price of the last thirty minutes that had served for decades. Orders are collected in a dedicated window and matched at one price, the mechanism already familiar from the pre open call auction at the other end of the session.

The derivatives leg did not change with it. The exchange's files show the futures close equal to the daily settlement price on every non expiry session in this sample, including those after the auction went live. So a closing basis measured after that date joins an auction determined cash price to a settlement determined futures price, where before it joined two averages. Treating a series that spans the date as one homogeneous sample pools two definitions of one leg.

That is an instruction, not a finding. Here the mean near month basis was 0.2200 per cent across the 124 sessions of 2026 before the change and 0.3315 per cent across the 31 after, with dispersion of 0.1796 and 0.2773. That comparison settles nothing. The post period runs 31 sessions and falls almost entirely in a month whose own mean carry across the sample is 6.96 per cent, so the two effects cannot be separated with this much data. The definition changed; the sample is too short to say anything else. That is the discipline set out in the guide to benchmarking against the right null.

A widening basis is a funding signal before it is a mood

This is the failure mode generic pages omit entirely. A basis widens, someone calls it bullish positioning, and the reasoning stops. Run through what actually widens a basis and direction is nowhere near the top of the list.

Funding got more expensive. The upper bound is set by the marginal participant's own cost of money, which moves with liquidity conditions, quarter ends, and anything that makes balance sheet scarce. The basis moves with it and nobody formed a view.

The window contains fewer distributions. Roll from a period with heavy ex dates into one with few and the negative leg shrinks, so the basis widens. A calendar effect, visible in the monthly table above.

The enforcing trade got harder. Margins rose, a limit bound, or the cash basket became more expensive to assemble. The residual widens because the band widens, and a wider band is not a direction.

The period got longer. A basis in points grows with time to expiry for a trivial reason. Comparing one a week from expiry with one four weeks out, without annualising, compares two different quantities.

The diagnostic is three steps. Annualise the basis over the actual days to expiry taken from the contract. Compare it with a funding rate rather than with zero. Check the distribution calendar for the window. If the number is still unexplained, the residual is worth thinking about, and the thinking starts with the cost and availability of the arbitrage.

What the basis is actually for

Used correctly the basis measures the cost and the availability of financing, one of the few things about a market genuinely observable from a screen. An implied carry sitting persistently above every published short rate says what capital actually costs the people willing to carry the position. One that goes negative says the short side of the arbitrage is unavailable.

Used as a direction gauge it is worse than useless, because it is confidently wrong in a stable direction. The forward is above spot almost all the time for reasons unconnected to the index going up, and treating that as a standing bullish signal builds a permanent bias into every chart carrying a futures price. The discipline is the ordinary one: know which quantity you are looking at, what it is defined against, and which of its parts you observe rather than infer.