How to Read Derivatives Market Data: One Figure at a Time
Every published derivatives figure counts one thing and one thing only, and most misreadings come from treating two of them as the same quantity. Below, five figures are read separately: what each one counts, what it can never be turned into, and what has to sit beside it before it means anything.
In any table of derivatives statistics the trouble starts before a single number has been read. The columns look like the columns in a table of sales or deposits. The columns carry a familiar unit. Several of them are printed in rupees crore. And the arithmetic ordinarily run on such a table, the growth rate, the ratio, the share of the total, all of it works perfectly happily and produces answers that mean nothing whatsoever.
The reason is not that derivatives are exotic. A derivatives table quietly mixes three different kinds of counting in adjacent columns, and it gives no warning when it changes from one to another. Consider the tables that are already trusted every day. A shop's stock sheet counts what is sitting on the shelves this evening. Its sales book counts what walked out of the door during the month. Its till roll counts money that actually changed hands. Those three sheets have different names and live in different drawers, and nobody at that shop would divide the stock sheet by the sales book and call the answer anything. A derivatives table puts all three kinds of counting in one grid, under headers that sound alike.
Two companies carry every figure below. Chitrakoot Cements Limited, an invented manufacturer, has agreed to hand over 7.20 per cent a year, a rate that never moves for as long as the agreement runs, and to take in whatever the floating benchmarkA published rate that a contract points at, so both sides read one agreed number off the same source instead of arguing about what the rate is. reads in exchange. Saranga Capital Limited, invented for the other side, stands on that same agreement, paying the benchmark and taking in the fixed rate. One notional of Rs~1,000 crore sits under both legs. The benchmark stays unnamed throughout, so no reading of any real benchmark is implied.
A single reading for a single period is all that the worked case supplies. One reading fixes what each figure means and never what any figure currently says, and at the point where a picture would normally be assembled, one reading will not support one.
A table shows the outstanding notional in a rate derivatives market rising sharply across a year. What has actually risen?
What is the one question to ask before reading any figure at all?
The question is: what does this row count? Not what it is called. The name was chosen by whoever built the table and may be a house term. Not how large it is. Size without a subject is not information. Just what, exactly, is being counted in each cell.
A row in a derivatives table has only three possible answers, and almost every misreading in this subject is a confusion between two of them. One row counts agreements, meaning how many separate arrangements were standing. Another counts multipliers, meaning the notionals written into those arrangements added together. A third counts money that moved, meaning amounts that genuinely left one bank account and arrived in another. The first answer has no currency in it at all. The second is printed in rupees and describes nothing that anybody paid. Only the third is money.
Once that question is answered, three more items go into the notes before the size of anything is considered. The unit, written exactly as the table writes it. Rupees crore and rupees lakh sitting in neighbouring tables have wrecked more comparisons than any conceptual error ever has. The date or the period, and which of the two it is. A figure taken on the last day of March and a figure covering the whole of the year to March are different animals wearing the same coat. And the name of whoever compiled the row. The choices that produced the number were theirs, and they will have written those choices down somewhere near the table.
The size of the figure is the last thing read, not the first. The inversion feels wrong for about a week and then becomes automatic. A large number pulls at the attention in a way a footnote never does. The discipline is to refuse that pull long enough to establish what the number counts.
A row headed in rupees crore carries a very large figure. What goes into the notes before its size is read?
Reading one: what is an outstanding notional a total of?
An outstanding notional is the sum of the notionals written into a set of agreements that had not yet ended at a given date. Every word in that sentence is doing work. A sum is an addition of things. The things added are notionals, and a notional is a multiplier rather than an amount. A date defines the set, and a date fixes the figure at a single instant rather than across a stretch of calendar.
Adding multipliers together produces a multiplier, not a size. Chitrakoot Cements and Saranga Capital have Rs~1,000 crore written into their agreement. Put ten similar agreements alongside and a compiler will report Rs~10,000 crore. Each of the eleven arrangements settles for an amount that depends on the gap between two rates on that particular agreement, and the sum has thrown all twenty-two rates away before anybody sees it. Nothing about that figure, then, says what any of the eleven will settle for.
Here is the property that catches people who are otherwise careful. An outstanding notional can rise sharply without one extra rupee of settlement happening anywhere in the market. Two parties sign a new agreement in March. The total goes up by the notional on it. Whether anything is ever paid under that agreement, and how much, depends on where the benchmark reads on each of its reset dates over the years ahead, and none of that has happened yet. The figure moved on the day of signing. The only thing that changed that day was a piece of paper.
An outstanding notional is not a useless figure for that, and leaving it there would be lazy. The total says how much reference amount contracts have been written against, and that is a real and meaningful measure of how much of this activity is going on. Read as a measure of activity it is honest. Read as an amount at risk it is not even wrong, it is a category error, like answering how heavy is a library with a count of its volumes.
Somebody hands the analyst an outstanding notional total and asks what the market has at stake. What factor turns the one figure into the other?
The same arrangement, read into both totals
Work it once on the arrangement in hand and the abstraction disappears. Chitrakoot Cements pays 7.20 per cent a year on a notional of Rs~1,000 crore over one full first period, so its leg comes to Rs~72.00 crore gross. Saranga Capital pays the benchmark. The benchmark read 6.00 per cent a year for that first period, giving Rs~60.00 crore gross. The day count fractionThe rule the two sides agree on for turning a stretch of calendar into the multiplier a rate is applied through, settled in the document long before anybody computes a payment. is 1.0000, one whole period, so nothing is scaled down. The two legs difference to a net of Rs~12.00 crore, and Chitrakoot Cements hands that across.
Now read the same arrangement into each of the two totals. Into an outstanding notional total, it contributes Rs~1,000 crore. Into a total of money that moved in the first period, it contributes Rs~12.00 crore. One agreement, two totals, and the two contributions differ by a factor of 83.33. Divide Rs~1,000 crore by Rs~12.00 crore and there it is. Put the other way round, the settlementThe moment money genuinely passes between two sides, as distinct from the moment an amount is worked out on paper. is 1.2 per cent of the notional it was computed from. For every Rs~100/- of notional that is Rs~1.20/-, one hundred and twenty paise.
| What this one arrangement contributes | To which total | Figure |
|---|---|---|
| The notional written into it | Outstanding notional at any date while it runs | Rs~1,000 crore |
| The fixed leg, gross | Neither total on its own, it is working | Rs~72.00 crore |
| The floating leg, gross | Neither total on its own, it is working | Rs~60.00 crore |
| The net that actually moves in period one | Money that moved, first period | Rs~12.00 crore |
| The notional again, in the period it was agreed | Traded volume, once only | Rs~1,000 crore |
| The notional again, in any later period | Traded volume, later periods | nil |
| Anything at all in period two or after | No total, because no reading exists | not available |
The last row repays attention rather than a skip past it. There is no reading for the benchmark in any period after the first, so no figure can be attached to period two, and none is. The empty row is not an oversight. A single reading supports no figure for period two, and nothing goes into a cell where nothing was observed.
Reading two: what does a traded volume figure count, and over what?
A traded volume figure counts what was newly agreed between two dates. Traded volume is a flow, in the same sense that the shop's sales book is a flow: it fills up over the month and starts again empty. The outstanding notional is a stock, like the stock sheet. A stock sheet describes a state on one evening and says nothing about how it got there.
The two are printed in the same unit, sit in the same table, often in adjacent columns, and cannot be compared with one another. One answers what exists at this moment. The other answers what happened lately. The numerator and the denominator are answers to different questions, so a ratio between them has nothing behind it, however readily any spreadsheet in the world computes one.
The Chitrakoot Cements arrangement makes this concrete on a single row. In the period it was agreed, it goes into a traded volume figure once, at Rs~1,000 crore. In every period after that, nothing is newly agreed, so it contributes nothing at all to traded volume. Yet it stands in the outstanding notional total at Rs~1,000 crore on every single date until it ends. One arrangement, firing once into one figure and standing permanently in the other.
The trap that follows from a stock and a flow sharing a table is quieter than the ratio, and it costs more. Somebody compares this period against last period. Perfectly reasonable, except that they have to check both figures are the same kind before they compare, and the table will not tell them. Compare a stock against a flow across two periods and the resulting growth rate is the sum of two unrelated movements, presented as one.
Two figures sit in the same table and carry the same unit. One is taken at a single date, the other counts a whole year. What is the pair good for?
Reading three: how can one set of agreements produce two different totals?
Every privately agreed arrangement has two sides. Chitrakoot Cements has one. Saranga Capital has the other. If both of them report what they hold, and a compiler adds up everything reported, that single arrangement has just contributed its notional twice.
Whether a published total counts an arrangement once or from each of its two sides is a choice made by whoever compiled it, and it changes the total without a single agreement being written or ended. This is why two credible compilations of the same market can differ by roughly half and both be entirely correct. Neither has made an error. The two compilers answered slightly different questions and said so.
The words said so are the part that matters. A compiler states the basis: which reporters are included, whether each arrangement is counted once or twice, what is excluded. The basis statement sits beside the table, in the notes, in the methodology document, in the header of the file. A reader who has not found the basis statement does not yet know what they are reading, however carefully they have read the number itself.
A related pair appears in the very same tables. One measure is gross: every arrangement counted, nothing set against anything. The other is taken after arrangements between the same two counterpartyEither of the two named sides to an agreement, each of whom is the other one's counterparty and each of whom carries the risk that the other fails to perform. names have been set against each other. A pair who have written offsetting arrangements against one another then appear once, at the residual, rather than at both full notionals. The netted figure is smaller. Both describe the identical set of agreements. Neither is more correct than the other, and which one is wanted depends entirely on the question brought to the table.
Two published totals describing the same market differ by roughly half. What is the first thing to go and check?
Reading four: what does a quoted rate say on its own?
A rate is not a quantity of anything. A rate is an instruction for producing a quantity, once the amount it applies to and the stretch of time have both been supplied. The reading discipline for a rate is therefore different in kind from the discipline for the three totals above, and a rate lifted out of its surroundings loses more than a total does.
Three things travel with a quoted rate, and without all three the number is worth nothing at all. The maturity it was quoted for, meaning the length of time the quotation applies to. The date and the time it was observed. Rates move during a day, and a quotation without a timestamp is a quotation without an identity. And the source that observed it. Two sources observing at the same instant can print different numbers and both be reporting honestly.
Dropping any one of the three has consequences. Without the maturity, there is no telling whether the rate covers a short stretch of time or a long one, and those are different numbers for reasons that have nothing to do with this subject. Without the date, there is no telling whether it is current, and a rate that was true three quarters ago is not a rate. Without the source, nobody can go back and check, and nobody downstream can check either.
A quotation for one maturity also says nothing whatever about any other maturity. How rates sit across different lengths of time is a subject with its own machinery, covered separately. The term structureThe shape formed when rates for many different lengths of time are set out together, which is studied on its own terms in the fixed income material. is met elsewhere as a whole object, and one point is not a shape, however carefully it is plotted.
One more habit is worth building here. Quotations move in small units and are usually discussed in basis pointOne hundredth of one percentage point, which is the unit rate quotations are usually moved and discussed in. terms rather than in whole percentages. Write a movement down without its unit and the result is a hundred times wrong. Write the unit next to the movement every time, in the same stroke.
A swap rate arrives with nothing else at all alongside it. What can be done with it?
Four figures have now been read one at a time. What would they normally be assembled into?
Reading five: what is the figure most wanted, and why is it absent?
Market data of this kind is normally read across a whole range of maturities at once. The reading calls for a rate for a short stretch, a rate for a medium stretch and a rate for a long one, each observed on the same day from the same source, and the three are read together because the relationship between them carries more than any one of them does.
The reading across maturities cannot be made from a single observation, and the reason is worth stating plainly. The record behind this material holds one fixed rate at one maturity and one benchmark reading for one period, and no series of any kind. There is no schedule of what the benchmark reads at later dates. There is no set of rates at other maturities. So there is nothing to plot, nothing to compare across lengths of time, and nothing that could be turned into a present value for the agreement.
Three things would have to exist before the assembling could honestly be done. A dated schedule of readings, at each maturity intended for use, from a source that can be named, compiled on a basis that source has written down. All three have to be met at once. Meeting two of them produces a picture that looks finished and cannot be checked. Somebody acts on a picture like that, and a picture like that is worse than none at all.
The five readings stand separately, and no single picture is assembled from them. Stopping there is the discipline rather than a shortcoming. A reader who takes away only that one habit, that figures get combined when the inputs support combining them and are left apart when they do not, has taken away the part of this that transfers to every table they will ever open.
The failure: the growth sentence
The most expensive sentence in this subject is written every year by people who have read the figure correctly and the caption carelessly. Outstanding notional rises between two dates. The next sentence says the amount at stake in this market has risen by the same proportion.
Watch what that sentence has quietly done. The sentence does two things, and only one of them is obvious. First, it converted a sum of multipliers into money, and no such conversion is possible. Each agreement inside the total settles for an amount that depends on gaps between rates, and the total carries no rates in it. Second, and this one nobody catches, it assumed the two dates were compiled on the same basis. A change in whether arrangements between the same two names are set against each other, or a change in which reporters are included, moves the total without one agreement being written or ended anywhere.
The cost is that the sentence gets repeated downstream by people who never saw the table, and by then no correction can reach it. It arrives in a summary, then in a note, then in a presentation, each time a little further from the footnote that would have stopped it. Somebody eventually makes a decision against a number that was never a measure of what they thought it measured.
How somebody actually uses these five readings on a working day
Take a treasury analyst at a mid-sized manufacturer who has been asked, on a Tuesday afternoon, whether the company should be worried that this activity is growing. The honest version of the work is small and takes twenty minutes.
The analyst opens the published table. They establish first which rows are stocks and which are flows, writing it beside each row in their own notes rather than trusting memory. Then they find the basis statement and read whether arrangements are counted once or from both sides, and whether the set of reporters changed between the two dates being compared. Only then do they look at the size.
The analyst hands back a sentence about activity and never a sentence about amounts at risk. A sentence about amounts at risk has no support in anything they read. Something like: the reference amount that contracts have been written against grew over the year, compiled on the same basis at both dates, and nothing in this table says what any of it will settle for.
A lender reading the same table does something narrower still. The lender is not trying to size a market. A lender wants to know whether a borrower in front of them is carrying arrangements the accounts do not show clearly, and no market aggregate answers that. The borrower's own schedule of arrangements answers it, and that schedule is a different document entirely. The market table tells the lender what the environment looks like and tells them nothing whatever about the name on their desk.
And a journalist, who has the shortest deadline of the three, has the most useful rule available: quote the figure with the words the compiler used for it, and never with a paraphrase. A paraphrase is where the stock quietly becomes a flow and the multiplier quietly becomes money.
What is done instead of copying a number?
Read the figure at its publisher, with its date beside it. Reading it anywhere else costs the date, and the date is most of what makes a figure usable.
Each kind of figure sits somewhere specific. Publication requirements for outstanding privately agreed arrangements are set by the Reserve Bank of India at rbi.org.in. Publication requirements for exchange traded contractA contract bought and sold through an organised marketplace on standard terms, rather than settled privately between two named sides. activity are set by the Securities and Exchange Board of India (SEBI) at sebi.gov.in. Cross border statistics on privately agreed arrangements are published by the Bank for International Settlements at bis.org, and the sensible move there is to open the release and read the figure in the release.
A number copied away from its publisher arrives carrying the date of the copy instead of its own. A reader who copies it out inherits a staleness they have no way of seeing, because the copy looks equally confident whether the figure was read this morning or eighteen months ago. Naming the source instead lets the date travel with the number automatically. No other mechanism keeps working after everybody involved has forgotten the conversation.
Why does naming a publisher beat copying the publisher's figure?
What does a careful reading of all five still not tell?
Suppose everything above has been done correctly. How much reference amount contracts have been written against is known. How much was newly agreed lately is known. The basis each total was compiled on is known, and the statement is there to point at. There is one rate, for one maturity, observed on one date, from a named source.
None of it says whether any particular party ought to be in any particular arrangement; that is a separate question. Market data describes an environment and never describes a position. What would have to be known first sits entirely outside every table discussed here: what that party already owes and on what basis, what its own receipts and payments do when a benchmark reading moves, and how much of a movement it could absorb without difficulty.
None of that appears in market data of any kind, at any level of detail, from any publisher. A person who has read the five figures perfectly has read the weather. Whether to put on a coat is a question about the person, and it is answered from their own records rather than from the forecast.
A colleague reads a quoted swap rate as the market's expectation of where rates are going. What has gone wrong?
Which authority sets what has to be published?
Publication requirements for outstanding privately agreed arrangements are set by the Reserve Bank of India, rbi.org.in.
Publication requirements for exchange traded contracts are set by SEBI, sebi.gov.in. Either requirement can change without notice, and the version in force sits at the address beside it.
Several things sit just outside this reading routine, and knowing which side of the line each falls on is part of using it.
Where any current figure actually stands is a matter for its publisher, read at the publisher's own table with its own date beside it.
The shape of rates across different lengths of time belongs to the fixed income material, where a curve is constructed, interpolated and fitted.
A notional, a leg and a reset are each taught in full elsewhere in this material.
Valuing an arrangement today is a separate subject. Valuing one needs a schedule of benchmark readings, and a single reading is not a schedule.
How a benchmark reading is produced and administered is set by the Reserve Bank of India, rbi.org.in.
Keeping a market observation from turning into an instruction to act is worked through in full further along in this material.
Where each of these figures is published
| Named here | What is read there | Site | Confirmed |
|---|---|---|---|
| Reserve Bank of India | What has to be published about outstanding privately agreed arrangements, and by whom | rbi.org.in | 28 August 2026 |
| SEBI | What has to be published about exchange traded contracts, and by whom | sebi.gov.in | 28 August 2026 |
| Bank for International Settlements | Where cross border statistics on privately agreed arrangements are released, to be read in the release with the date printed on it | bis.org | 28 August 2026 |
Chitrakoot Cements Limited and Saranga Capital Limited are invented.
Educational material. Not advice on any investment, tax, budget or market position.