Structure and liquidity

The problem of someone who must buy a thousand contracts

You buy one contract whenever you feel like it. It is so obvious that nobody gives it a second thought, and yet it is the starting point of this entire category. Your order always finds someone on the other side, because it is tiny next to what is offered at the displayed price. The market never registers that you were there, and that invisibility feels normal to you because you have never known anything else.

Change one single thing and it all breaks. Suppose you have to buy a thousand, today, on an instrument whose book shows sixty contracts for sale at the best price. You send the order in one block. It eats the sixty, then the seventy one tick up, then the ninety above that, and it keeps climbing the ladder until it is filled. Every extra contract costs you more than the last, and you are the one who pushed up the price you are paying.

Look at the figure and add the column up, it says more than any argument. Everything offered for sale across the six visible levels does not add up to a thousand contracts. The order therefore cannot be filled at the displayed price under any circumstances: it can only push price up until enough sellers show themselves, or wait for them to arrive on their own.

Nothing about this constraint is secret and none of it requires an agreement. It follows from an accounting fact the order flow lesson demonstrates in detail: every transaction has exactly one buyer and one seller, for the same quantity, at the same instant. Wanting to buy a thousand contracts means wanting to find people willing to sell a thousand. Nothing guarantees they exist today, at that price, or even within the hour.

⚠️ Keep the wording, because everything else depends on it. A large participant's problem is not to predict price, it is to find counterparty. Those are two different jobs, and the second is the one that leaves visible marks on your chart.

What “working an order” actually means

Since the order will not go through in one block, it has to be worked. The verb belongs to the trade, and it covers three decisions: how many pieces to cut it into, how fast to send them, and where to sit so that counterparty comes to you instead of you having to chase it.

The first person to put a name on what all that manoeuvring costs was André Perold, in 1988, in a paper that became a classic, The implementation shortfall: paper versus reality. His idea is brutally simple. Compare two portfolios: the paper one, where everything fills at the price the screen showed the moment the decision was made, and the real one. The gap between them is the cost of execution, and it almost always dwarfs the commissions everybody talks about.

Twelve years later, Robert Almgren and Neil Chriss formalised the central trade-off of the job. Go fast and you pay impact: your own demand pushes price against you. Go slow and you pay timing risk: the market can leave without you while you wait, and you end up buying far higher for a reason that has nothing to do with your order. No setting cancels both at once, only a compromise to pick and to live with.

The third lever, the one that concerns us here, is place. An order sitting in the right spot gets filled without pushing anything. An order chasing price pays twice, once in impact and once in regret. The question of where to sit in order to be filled is not a specialist's refinement, it is what decides the bill, and it points at places you can see too.

On centralised markets, part of that work is tooled directly by the exchange. A hidden quantity order, also called an iceberg, displays only a fraction of its size and reloads as it is filled. Nothing clandestine about it: the function is written into the market rules, open to anyone who asks, and its very existence proves that hiding your size is an ordinary need, not a trick. Working an order is not a matter of being clever, it is a matter of being constrained: a trader with a small position never has to think about any of this, which is why none of it appears in beginner material, and why the first encounter with the idea usually arrives dressed up as a secret.

Market impact, and what it really costs

Market impact is the price displacement caused by your own order. It splits into two pieces that are worth keeping apart. One is temporary: you emptied the book, it refills in seconds or minutes, and price comes back to where it started. The other is permanent: the market read your flow as information, concluded that somebody knew something, and revised its idea of the right price.

Albert Kyle gave the idea its theoretical frame in 1985, in an Econometrica paper that remains the reference of the field. He models a market as having a measurable depth, meaning a quantity it can absorb per unit of price displacement. That depth is not infinite, it varies from hour to hour, and it is the real resource large orders compete over.

What measurement later showed is more interesting than the model. The extra cost per contract does not track the size of the order, it roughly tracks its square root, relative to the day's usual volume. That regularity, spotted as early as the nineteen eighties and confirmed on ever larger datasets since, is known as the square root law of market impact. Concretely, doubling the share of volume taken does not double the extra cost per contract, it multiplies it by about one point four. Multiplying that share by a hundred only multiplies the cost by ten. The figure sets the two curves side by side.

The practical consequence fits in two sentences. Quadrupling the size does not quadruple the unit cost, it only doubles it, which is good news for the very large. On the other hand a tiny share of the volume already costs far more than its proportion would suggest, which is bad news for very nearly everyone. The curve is concave, and a concave curve punishes the first steps hardest.

⚠️ None of these numbers apply to you directly, and that is precisely the point of the lesson. An ordinary retail trader produces no measurable impact on a heavily traded instrument, the transaction vanishes into the noise of the session. On a thinly traded one the same transaction shows, and the immunity leaves along with the depth. That immunity is a privilege rather than a given, and further down we will see it comes with an advantage money cannot buy.

The question they ask is not yours

Put yourself for a minute inside the head of whoever has to place those thousand contracts before the session closes. Open the same chart they have, the same candles, the same levels. What you look for on it and what they look for have nothing in common.

Your question is a question of direction: is it going up, is it going down, do I get in now. It is perfectly legitimate, it fills most of the trading literature ever written, and it assumes price is the subject of the day. Theirs is a question of address: where, on this screen, are there enough people willing to sell to me. Price is not their subject, it is their constraint. They will buy lower if they can, but they will buy either way, because the decision to buy was taken elsewhere, often by someone else, and often the day before.

That difference explains a behaviour that looks absurd from your chair: a very large buyer letting price drift down without lifting a finger, or resting orders well below the market and waiting. They are not sabotaging themselves, they are settling where counterparty will turn up in size, because it is the only place on the chart where they can be filled without blowing up their own bill.

⛔ Beware of the shortcut almost every piece of writing on this subject takes. They do not need to push price down there, they need price to go there, which is a different thing and asks nothing of them. Deliberately forcing a move in order to set off other people's orders has a name of its own, market manipulation: it is banned on every regulated venue, it is prosecuted, and it has nothing to do with the ordinary execution of an oversized order. A chart never lets you tell the two apart, which is reason enough not to rule on it in the chart's place.

Where does that counterparty pile up? In places you can see just as clearly as they can, and that is what makes this reading usable rather than merely fascinating: under an obvious low, above an obvious high, at the edges of whatever everybody has in front of them. Those clusters are hidden from nobody, they are the mechanical consequence of everybody having learnt the same rule from the same books. How thick they really are, what happens when price reaches them and what you should do about your own orders fill an entire lesson, the one on liquidity and stops. This one stops at the principle.

How a large order is executed, in public

The methods exist, they have names, brokers sell them from a catalogue and describe them in their documentation. Nothing that follows is confidential, and simply knowing it defuses half the mysterious stories written on the subject.

The simplest one slices the order through time, in equal pieces over a chosen duration. It is known as time weighted average price slicing. Its virtue is its simplicity, its flaw is its regularity: an attentive observer eventually recognises the rhythm and steps in front of it.

The next one slices through volume. The algorithm follows the session's usual profile, buying more when the market trades heavily and going quiet when it dozes. That is volume weighted average price slicing. A blunter variant simply fixes a share of volume, for instance never accounting for more than a tenth of what changes hands, and lets the order run for as long as it takes.

The last family targets Perold's measured cost directly, trading impact off against timing risk continuously according to current volatility. It is the Almgren and Chriss trade-off put into code. Alongside these slicing schemes sit block trades, negotiated between two counterparties who find each other away from the book, then reported to the market. The European framework that came fully into force in 2018 wraps all of it in a best execution obligation, which requires the intermediary to demonstrate that it sought the best possible result for its client.

Keep the logic rather than the names. Every one of these methods does the same thing: it trades discretion for time. They therefore assume the order has time, and that assumption collapses on a regular schedule, at a contract's expiry, at an index close, at a portfolio rebalancing date. Those are the moments that produce moves which seem to come from nowhere, and they only seem to for anyone who does not know the calendar.

Your structural advantage, the one they cannot buy

Here is the passage the conspiracy version of this story never tells, because it takes all the flavour out of the victim's tale. You hold, over those participants, an advantage no amount of capital buys back, and it follows from exactly the same constraints described since the beginning of this lesson.

You get in and out whenever you want. Your position is built in a second and unwound in a second, give or take the spread, pushing nothing: your round trip costs you that gap between the buying price and the selling price, twice, and nothing else. They take a day to get in and a day to get out, and they pay impact on both sides. An idea only worth taking for twenty minutes is available to you and forbidden to them: by the time they finish getting in, the opportunity has gone.

You are under no obligation to be present. It is the most underrated of the three. A manager reports against a deadline, against a benchmark, under a mandate that often forbids sitting outside the market at all. Doing nothing costs them their job. Doing nothing costs you strictly nothing: no commission, no spread, no capital tied up.

Your size finally makes you invisible. Nobody sees you coming, nobody positions ahead of you, no observer reconstructs your intention from your flow. What the large spend fortunes concealing, you have for free, without thinking about it, since your very first order.

⚠️ An advantage is not a result. Those three freedoms pay nothing as long as they go unused, and the most common mistake is precisely to squander them: forcing a position every day, holding a trade past the idea that justified it, feeling obliged to have a view on everything. Imposing a fund manager's constraints on yourself without having either their means or their reasons means paying the price of their job without collecting a single one of its advantages.

“Smart money”: what it means, and what it does not

The phrase is everywhere, it has no stable definition, and that is the whole problem. Taken seriously, it denotes participants whose size imposes the constraints described above: funds, market makers, corporate treasuries, discretionary mandates. Nothing else. The word smart in it is an accident of vocabulary, not a demonstrated property.

Academic literature, for its part, never had a category called smart money. Larry Harris, in Trading and Exchanges published in 2003, still the reference textbook on market mechanics, sorts participants by the reason they trade: those who trade because they have a need, those who trade because they believe they know something, those who supply counterparty in exchange for a fee. Not one of those three categories assumes second sight.

What the phrase does not mean is worth listing, because every misreading costs somebody money. It does not denote people who know the future: a very large participant is wrong on a regular basis, and their size stops them getting out quickly when they are. It does not denote a coordinated group: large participants are first and foremost each other's competitors, and they take money off each other every day. It does not denote anybody aiming at you personally: your order is invisible at their scale, which is exactly the advantage of the previous section.

The test that settles it is borrowed from Karl Popper, who argued in 1934 that a claim is only worth something if you can imagine an observation capable of contradicting it. Apply it as it stands, adding nothing. An explanation telling you why an order needs counterparty and where that counterparty piles up makes a prediction: you can go and log a hundred lows and count. An explanation telling you that “they” came for your stop makes none, since it stays true whether price turns around or keeps falling.

⛔ The second family of explanations is comfortable, because it converts every loss into an injustice and excuses you from looking at your own decision. That is also why it sells so well, with an abundant vocabulary and a great many capital letters. The genuinely useful part of this whole field is small, mechanical, checkable, and it fits inside the two lessons that follow, the one on market structure and the one on liquidity and stops.

What this reading does not prove, and how to measure what it claims

An honest school says where its teaching stops. This family of readings explains a constraint, not a schedule. It announces neither which level will be visited nor when, it identifies nobody, and a chart carries no information whatsoever about who bought. Everything readable on it is a consequence, never a signature.

⛔ Two sentences deserve to be refused on sight, including when you catch yourself writing them. The first assumes an agreement: large participants are each other's competitors first, the constraint described here applies to each of them separately, and no coordination whatsoever is needed for the result to be the one you observe. The second turns a constraint into a timetable: knowing where counterparty piles up does not mean an order will come for it, not today, not this week. Most of the time nobody comes, and nothing on the chart will have warned you.

Nor does anything prove that a very large order lies behind every fast move. A headline landing, a market maker pulling their prices for thirty seconds, a string of automated orders setting each other off all produce exactly the same candle. The result on the screen is identical, the explanation is different, and candles do not let you decide.

The exercise that makes you touch the constraint takes a quarter of an hour and asks for nothing but your platform. Open the depth of market on a heavily traded instrument, the panel showing the quantity available at each price, and write down two numbers: the quantity offered at the best price, and the total displayed across the first five levels. Do it again on an instrument almost nobody trades, then a third time on the first one, at a dead hour. Three readings, six numbers, fifteen minutes.

Compare them, and look at what you have just done: you measured Kyle's depth by hand, on your own screen. The three readings have nothing in common with one another, and that is the entire result of the exercise. This quantity is small, it is public, it changes from one instrument to the next and from one hour to the next, and it is what decides what price does when an order too large for it turns up. If your platform shows no depth at all, which is the case on many contracts for difference, take the volume of a one minute candle at the same three moments instead: the measurement is cruder, the conclusion does not change.