Market Manipulation: The Patterns Surveillance Looks For

Three shapes appear in a record: activity crowding into the minutes a price is being fixed, trades that end with everybody holding exactly what they held, and orders placed with no intention of letting them execute. The definition of manipulation belongs to the authority, and every finding turns on it.

Every one of these shapes matters in the same place, and it is worth naming that place before anything else. Clearing machinery runs on determined numbers that nobody inside it chose, and the heaviest of them is the closing price against which every position still standing gets marked. Touching that number reaches into what gets collected from every position in the contract that same evening. Which is why the watching sits next to the collateral rather than off in some corner of its own.

So what exactly does a record describe?

Three descriptions, and nothing beyond them. A certain kind of activity has a look to it once it is laid out as entries in a record, at the resolution a record actually offers: what happened, when, in what size, and how it sits next to other things. Most people who work near a market never see that laid out at all, which is what makes the describable part genuinely useful.

The line between something and nothing is a line data does not draw. The Securities and Exchange Board of India (SEBI), at sebi.gov.in, holds the definition of manipulation. SEBI holds the standard anything has to be established to. SEBI holds everything that follows once a finding is made. Each of those changes over time, and they are covered separately, in the material on rules.

A reader is tempted to take that division as legal throat-clearing before the interesting part. The division is the opposite. The reason it matters to a reader rather than to a lawyer is this: a shape is something anybody with the record in front of them can see, and whether that shape was anything at all is a conclusion that needs evidence the record does not contain. The two live on different sides of a wall. A reader can become extremely good at seeing shapes and get no closer to the second thing, and believing otherwise amounts to a confidence nobody has earned.

Think of a security camera above a shop door. The camera shows a man leaving the shop at three in the morning carrying a box. The footage establishes that, and establishes it very well. Whether he was stealing, collecting his own property, or doing the shopkeeper a favour is not on the tape and never will be, however many cameras are added. Adding cameras improves the record. The gap is not a gap in the record, so no number of cameras closes it.

Why does any of this belong beside the subject of collateral?

Because of one number, and everything described here reaches an account through it. When the day closes, a settlement price is determined by a stated method, and each position still standing in that contract gets marked to it. Nobody chose that number as a matter of preference. The number came out of a procedure, and a procedure is precisely what makes it worth attacking: a number produced that way will be used by everybody, without argument, on the same evening.

Worked in this order, the sequence becomes concrete. One unit of the invented reference asset carries Rs 2,000.00/- of exposure. The agreed price a year out is Rs 2,130.00/-. Take the Rs 2,000.00/- and add Rs 130.00/- of financing at 6.50 per cent a year; no view about where the price is heading enters it. Holding the reference asset produces no income of any kind across the twelve months. Change that and the Rs 130.00/- of carry buried inside Rs 2,130.00/- changes with it. The point gets said out loud instead of assumed. On the day in question, the settlement price lands at Rs 2,050.00/-: take the agreed Rs 2,130.00/- and knock off what a 4.0 per cent adverse move costs on the exposure, namely Rs 80.00/-. Against that sits collateral of Rs 160.00/- a unit, struck at 8.0 per cent of that exposure and a figure chosen here for teaching.

Now suppose that determined price ends the day Rs 20.00/- lower than the stated method would have produced, so it prints at Rs 2,030.00/- instead of Rs 2,050.00/-. Take one hundredth of the Rs 2,000.00/- exposure and Rs 20.00/- is what falls out. Each long position in that contract is now called for an extra Rs 20.00/- a unit that nothing justifies, and the mirror of it lands as an extra Rs 20.00/- a unit reaching each short. Not the parties who dealt with whoever produced the shape. The revaluation runs off one number for all of them, and so every position in the contract is hit.

The three shapes are not offences against an abstraction; they move money between real accounts, through machinery that is working exactly as designed. The clearing corporation did nothing wrong. The clearing member did nothing wrong. The position holder on the losing end did nothing wrong either. Every part of the arrangement performed correctly on a number that was wrong when it arrived.

The bars carry the point better than a sentence does. Divide the exposure by what was lodged against it and 12.50 falls out. One unit is Rs 2,000.00/- of reference asset resting on Rs 160.00/- put up, so an ordinary adverse day of 4.0 per cent, worth Rs 80.00/-, already takes 50.0 per cent of the collateral and leaves Rs 80.00/- standing. Add the Rs 20.00/- that nobody earned and the payment called runs to Rs 100.00/-, a reading of 62.5 per cent against the collateral, with only Rs 60.00/- left standing. A quarter of an ordinary day's movement is what a one per cent distortion amounts to, and it lands on parties who did nothing at all that day.

Two figures that agree, and why the agreement is not chance

One numeral pairing above reads as chance. It is nothing of the kind. The Rs 20.00/- distortion amounts to one hundredth of the exposure. Measured instead against the Rs 160.00/- of collateral it eats into, the same Rs 20.00/- is 12.5 per cent. And the leverage on this unit is 12.50 times. The two figures print the same digits.

The agreement is forced arithmetic rather than a happy accident: the leverage multiple is exactly the conversion factor between a share of exposure and a share of collateral, so any move worth one per cent of the exposure is worth 12.50 times that as a share of what was put up. The two only look identical because the distortion here was set at exactly one per cent. Move it to 1.5 per cent and the same reading becomes 18.75 per cent of collateral, and nothing coincides any more. A reader who spots two figures agreeing and cannot say why will treat the agreement either as a check that passed or as a typing error. It is neither.

What does crowding around the moment of a fixing look like?

The going rate for tomatoes in a neighbourhood market gets fixed by whatever the last few carts are asking as the stalls come down, and every trader's week is squared up against that rate. A man with a hundred crates to square up walks in at the last minute and buys two crates at a high ask. He has bought almost nothing. He has moved the number that his hundred crates settle against. Every other trader in that market has just been repriced by a purchase of two crates.

In a record, the same thing has a shape. Activity gathers inside the window a determined price is drawn from, and it is entered by a party carrying a position that the determined price decides the value of. Two halves, and neither on its own is anything. Timing alone is ordinary and a position alone is ordinary; it is the two of them together that make a shape worth a second look.

Now the correction, which has to arrive in the same breath. It is at least as important as the shape. Somebody carrying a position has entirely ordinary reasons to act at the end of a day, and enormous numbers of people do. A desk is closing out before a weekend. A client instruction arrived at four o'clock. Their own risk limit is measured on a closing figure, so the close is when they must be flat. A treasury desk that must square its book daily is going to be busy in the last minutes of every single day of its working life, and its audit trailDated entries kept in sequence, showing who did what and at what hour. Reconstructing a day afterwards depends entirely on whether one was kept. will show that shape a few hundred times a year.

So the shape narrows the field and settles nothing. How the determined price is arrived at, and which window it draws on, come from SEBI's rules at sebi.gov.in. Both details matter enormously to whether any given afternoon even falls inside a window, and both belong to the authority.

How can a trade that moves nothing still be visible?

Two brothers with adjacent stalls pass the same sack of onions between them eleven times before noon. At the end of it, each brother has what he started with. The sack has not gone anywhere. But anybody standing at the end of the lane counting sacks changing hands will report a lively morning for onions, and the rate somebody quotes at midday will have that report inside it.

The answer is that it takes two records to see it, and that is the whole mechanism. A trade of this kind produces a perfectly ordinary printTrades leave a public mark once they are done: the price and the size go out for anybody at all to read. and adds perfectly ordinary volumeHow much changed hands over a stretch of time. Because other people's calculations feed on it, volume reaches past the two parties who traded.. Read the trade record on its own and there is nothing to find. Read the record of who is carrying what before and after, and the holdings are identical. Movement in one record beside stillness in the other is what makes the thing visible at all, and neither record shows it alone.

Here is the part that reaches back to the collateral. Prints and volume are not private between the two parties. Prints and volume are inputs to numbers other people read, and one of those numbers is the determined settlement price. So a trade that transferred nothing between its two participants can still change what everybody else in the contract is revalued at. The ugly bit of the arithmetic sits right there. A transfer of nothing between two parties can still produce a payment from thousands who were not involved.

An attributionTying an entry back to whoever produced it. Two accounts can behave identically on screen and belong to people with nothing whatever in common. question sits underneath all of this, and it is harder than it looks. Two accounts trading with each other establish nothing by themselves; most trades in any market are between two accounts. The position record shows whether the net effect on holdings was nil, and even then a nil effect has ordinary explanations, including a whole class of them that are simply administrative. Moving a position from one account to another for perfectly dull internal reasons leaves the holdings of the party as a whole unchanged, and it happens constantly.

What do orders that were never meant to trade look like?

Picture a small shop that pays six cousins to stand outside it in a queue every evening at six. Nobody in that queue intends to buy anything. The queue is not there for the shopkeeper's cousins; it is there for the strangers walking past, who read a queue as evidence and join it. The moment a stranger walks in, a cousin peels away.

In a record, the shape is orders entered and withdrawn in a pattern, at sizes and at distances from the going price that make execution unlikely, with activity running in the other direction at the same time. The shape manufactures the appearance of interest where there is none, and appearance is an input to other people's decisions. That last clause is the reason it can matter at all: the queue works only because somebody else reads it.

The bottom row of that figure is doing more work than the top one. Withdrawing an order is entirely ordinary, most withdrawn orders are exactly what they appear to be, and it is the pattern rather than the withdrawal that constitutes the shape. Anybody who has changed their mind about a price has withdrawn an order. Anybody whose system re-prices automatically as the market moves is entering and withdrawing all day without a human being involved. Counting withdrawals and calling the total suspicious would flag the most conventional participants in the market and nobody else.

The setting of a test is also where a false positiveA test firing where nothing is wrong. Any test sensitive enough to catch the real thing will do this, and will do it often. stops being a technicality. A test tuned tightly enough to catch a manufactured queue will also catch every automated system doing its job, and a test loosened until it stops doing that will catch nothing at all. Where the setting lands is not a question data can answer. The setting is a judgement about what an exchange watches for, and SEBI settles that at sebi.gov.in.

What can a pattern settle, and what will it never settle?

The limits of a pattern are worth stating plainly rather than gracefully. A pattern can establish that something happened. A pattern can establish when, to the second if the record is good. A pattern can establish size. And a pattern can establish how one entry sits in relation to others. The three shapes above are made of relations, and that is where nearly all of a pattern's power lies.

A pattern cannot establish why. Not because records are imperfect, and not because more data would fix it. A pattern is a question put to a record, and nothing inside that record ever answers it. The reason somebody acted exists in one place, which is with the party who acted. The reason was never in the data, so no amount of data recovers it. Inference is the dangerous word in this whole subject for exactly that reason: inferenceReasoning from something visible to something that is not. The step belongs to the person taking it rather than to the record, and it can go wrong while every entry behind it stays right. takes a reader from the entries to the reason, that step belongs to the person taking it rather than to the record, and that person is answerable for it in a way the record never is.

Two errors follow from forgetting that. Only one of them gets talked about, and they deserve equal weight. The first is treating a shape as a finding. The second, and it is just as costly, is treating the absence of a shape as evidence that nothing happened. Tests catch what they were built to catch. Something conducted in a way no test was built around leaves no shape, and a clean screen is a statement about the tests, not about the market. Neither the presence of a shape nor its absence carries the weight a reader instinctively wants to put on it.

What if a perfectly ordinary day makes one of these shapes?

It will. The arithmetic guarantees it: the shapes described above are produced constantly by activity that is entirely unremarkable, by people who have never given a moment's thought to any of this. Squaring a book daily makes the first shape. Moving positions between accounts under common ownership for administrative reasons makes the second. Re-pricing orders as the market moves makes the third. The shapes are not rare, they are not exotic, and being in one is not a statement about the party concerned.

The explanation that turns a shape back into an ordinary Tuesday exists in exactly one place, with the party who acted, and it decays. The decay is fast. Six months later that party remembers there was a reason and not what it was, and reconstructing it from their own records afterwards is an order of magnitude harder than noting it at the time. A contemporaneousWritten at the same moment as the thing it describes rather than recalled afterwards. Memory rewrites itself; a note made on the day does not. line saying why the thing was done, in a form somebody who was not there could follow, is worth more than any amount of careful reconstruction later.

Records behave one way over time: the cheapest moment to record a reason is the moment it exists, and every moment after that one is dearer. The same is true of a delivery note, a repair bill or a receipt for a wedding caterer. Nobody thinks of any of those as caution. Delivery notes and receipts are just what people who deal with records learn to keep.

Who has to do something about any of this, and what is the work?

Each shape starts somebody on a job, and naming those jobs shows the point of all this better than naming the desks they land on.

Somebody has to write the explanation down. A query arrives about activity in a closing window on a particular day, with a deadline attached. The work is not defending anybody; it is producing the reason, from records that either exist or do not. Where the audit trail carries an instruction received at four o'clock, the work takes twenty minutes. Where it carries only the trades, the work takes a fortnight and produces something weaker than the truth.

Somebody has to decide whether published volume can be trusted as a measure of interest. Anybody sizing a position against how easily it could be closed is leaning on volume, and volume is exactly the number that a trade moving nothing inflates. Knowing what volume is made of before leaning on it is a different skill from reading it.

Somebody has to price the same reference asset for a lender's collateral schedule. If a determined price feeds a valuation somewhere else, then a distorted determined price feeds a distorted valuation somewhere else, and the person on the other end of that chain never saw the market at all. A determined number has that quiet reach: it is used by people who have no idea it exists.

And somebody at home has to decide whether to be worried. A household that sold in the last ten minutes of a heavy day and later reads an account like this one will wonder. The honest answer is that the shape says nothing about them, that the shapes are common, and that a line in a notebook saying why they sold is worth more than the worry.

Who decides what any of this actually was?

The authority does. The whole answer is that short. Three things belong to SEBI at sebi.gov.in: what counts as manipulation, the standard anything has to be shown to, and what happens once a finding is made. Each one changes over time. The material on rules covers all of them separately.

Neither an exchange's test matching on somebody's activity, nor a description of a shape, is a finding of anything whatsoever. They are two different things that both feel like conclusions and are neither. A test matching is the beginning of a question. A description of a shape is vocabulary and nothing more. The distance between either of those and a finding is filled with argument, evidence and a hearing, and that distance is essentially the entire subject.

Principles for how cleared markets are watched across borders come from the International Organization of Securities Commissions (IOSCO), iosco.org. SEBI's own rendering of those principles is what applies in India, and SEBI's version, not the international one, governs anything happening here.

The questions a reader wants most at this point are answered nowhere in a record: how often one of these shapes turns out to be something, what became of whoever it attached to, what it cost them. Those three counts are produced by adjudication, one matter at a time, and they are published by the authority that adjudicated. A worked arrangement shows the machinery running; it never gets as far as anything going wrong inside it.

Then what is the question none of this answers?

A reader who has followed the three shapes arrives at one question above all others, and it is worth saying out loud instead of letting it sit there: was the thing I saw manipulation? No description of a shape answers that question. A shape is not the kind of thing that carries an answer to it.

Three things would have to be established first, and not one of them is a matter of data. First, what the definition actually requires; SEBI writes it at sebi.gov.in and it moves. Second, what evidence exists beyond the record, meaning the whole territory a record does not reach. Third, who has the standing to decide, a matter of who was given that power rather than who read the data most carefully. All three belong to the authority.

One further question sits alongside it. Rs 2,000.00/- of exposure standing on Rs 160.00/- put up prompts the thought of whether a position like this is worth holding at all. Settling that would take the job the position was opened to do, whatever is already held on the other side of it, what has been lodged so far, what more could be found by tomorrow morning, and how large a single day's demand could get at its most punishing rather than at its most usual. Those five inputs sit with the holder of the position and nowhere else. No worked arrangement supplies them, because a worked arrangement holds not one realised outcome and no series of them.

The relationship between a distorted price and the payments it produces is the one that genuinely can be worked, and the arithmetic above works it.

The labels, because mixing them is the commonest error here

The test for any figure above is what happened to it. Nothing happened to Rs 2,000.00/-, to Rs 2,130.00/- or to Rs 2,050.00/-. Each was agreed, quoted or worked out by a stated method, and each sits there as a PRICE. Something did happen to Rs 80.00/-. It left one account and reached another, and that makes it a PAYMENT. Where obligations pointing both ways have already been set against one another, whatever survives the cancelling is a NET.

One unit here stands on Rs 2,000.00/- of the reference asset. Multiplied out, that gives the EXPOSURE the collateral is struck against. A contract's face amount multiplied out instead gives a NOTIONAL, of which not one rupee ever moves. Both figures appear above, so one word about them. The Rs 2,000.00/- spot price and the Rs 2,000.00/- of exposure on one unit are the same numeral for the plainest of reasons: a single unit of the reference asset is precisely what a spot price prices.

The chain traced above stops at the collateral. The body standing in the middle has a balance sheet of its own, reserves of its own and people answerable for how it is run, and all three of those are covered separately.

The reference asset and the position holder who carries a single unit of it are invented.
Educational material. Not advice on any investment, tax, budget or market position.