Liquidity and smart money

Why the location of your stop is predictable

You buy after a pullback. The next question is always the same: where does the stop go? You look at your chart, you find the last low, the one visible without zooming, and you put your order just under it. That reasoning is correct. It is the only reasonable one, because below that low your reading is genuinely wrong.

The problem is not the reasoning, it is that the reasoning is shared. The beginner who opened an account last week, the manager following a written plan and the program executing a coded rule all look at the same low, on the same chart, with the same common sense. Nobody needs to coordinate to end up in the same place.

One property follows, and there is nothing mystical about it: the position of your stop is deducible without knowing you. Looking at the chart you are looking at is enough. It is an inference about shared reasoning, not a peek into your account, and it is the key to reading the rest of this lesson without falling into the persecution story.

A technical detail confirms it. A stop is not a limit order: it does not rest in the book, it is held aside by your broker or by the venue, and it exists for the market only in the second it triggers. The depth you can see therefore contains no stops at all, and what everybody can see is not your order, it is the low that made you place it.

A band a few points wide, not a line

Those orders do not land on the same price. One goes a point under the low, another three, a third five, because they read somewhere that you should let it breathe. The result is not a line, it is a band, a few points thick, whose density peaks just below the visible level and fades quickly as you go down.

That band has a property nothing else on the chart has. Once triggered, the orders inside it become market orders, meaning orders that execute at whatever price is available. A crowd of sellers in a hurry, gathered in one spot, with no price limit whatsoever. For anyone who has to buy in size without pushing the price up against themselves, no better address exists anywhere on the chart.

This is what separates a liquidity pocket from a plain support level. A support is a place where buyers may have stepped in. A liquidity pocket is a place where sellers will mechanically appear if price goes there, whether anybody wants it or not. The first is a hypothesis, the second is a consequence.

The width of the band depends on the instrument and on how agitated it currently is. In a quiet market it fits in two or three points, in a nervous one it spreads over ten. The useful reference is therefore not a fixed number, it is a fraction of the recent average range: what price covers in one ordinary candle is what you can lose for nothing.

The other side of the chart

Everything above flips over a visible high. The short seller puts a stop there, and that stop is a buy order: cutting a short means buying back, exactly as a stopped-out buyer has to sell. An exit from a loss is always an entry for somebody else.

A second population joins them, and it changes the nature of the place. Whoever is waiting for a break to enter also places a buy order just above the high. The two populations have nothing in common, one is cutting a loss and the other is opening a position full of hope, yet they park their orders at the same price, inside the same band a few points wide.

An obvious high therefore concentrates far more buy orders than people imagine, and that is a good part of the explanation for something everyone has lived through: the break of a level everybody watches leaves very fast, then comes straight back into the zone. The move was not carried by conviction, it was carried by an order-book mechanism that runs dry in minutes.

Spotting these places takes no indicator, only one question: which level does everybody see on this chart? Yesterday's high, the overnight low, the boundary of a three-day range, the high price the market turned away from last time. Those are the levels that count, not the ones you have to zoom in to find. A level only you can see concentrates nothing.

External and internal liquidity

Two words come up as soon as you dig into this subject, and they are useful provided you take them for what they are: a way of sorting the places where orders sit, not one more theory.

External liquidity is what lies beyond the visible extremes: above the last high, below the last low. That is where stops stack, for the reason seen earlier, and it is the largest pool and the easiest to locate. It takes a second to spot on any chart: both boundaries jump out, and what sits behind them needs no calculation.

Internal liquidity is everything resting between the two boundaries: the minor highs and lows you barely notice, the areas price crossed in one go leaving orders unfilled, the levels where a few positions built up without anyone making a reference of them. It is more scattered, smaller pool by pool, and far less obvious to place.

The practical difference fits in one sentence. A modest order fills in internal liquidity, quietly, because a handful of counterparties is enough. A large order cannot: it would exhaust those small pools before being filled, moving price along the way. It needs the big pool, the one that sits outside.

⚠️ This explains an observation that throws every beginner: a market that drifts inside an area for hours can suddenly go and fetch one of the two extremes, very fast, then come back inside as if nothing had happened. It did not go “nowhere”. It went to the only pool large enough to fill somebody.

Protected lows, and what gets swept all day long

The previous section says WHERE the two pools are. The most useful part is missing: which one counts, and how often each gets taken. Without that, the distinction stays a pretty map with no scale.

Not all lows are equal. Only one, at any moment, carries the trend: the last low higher than the previous one, the one whose break ends the series. That is what gets called the protected low. You have met it under another name already, in the Dow theory lesson: it is the last supporting low, the one whose break is the second of the two conditions of a turn.

The word “protected” is not decorative, and it points at something concrete. That low is defended, in the literal sense, by those who built their position on it. An institutional participant who accumulated below that level does not merely hope it holds: they have a mechanical reason to make it hold. If price goes through, their own reading is wrong, their entire position is under water, and what took days to build is worth nothing. So they add at that level, they absorb the selling that arrives, and that defence is what produces the clean bounces seen on every return. A protected low is not a place where price “should” bounce: it is a place where somebody has an interest in it bouncing, and the means to see to it.

It also explains the day it gives way. When that level goes, it is not merely a line crossed: it is a defence that failed, so institutional positions under water that have to be cut, at the same moment as retail stops trigger. Both flows run the same way, which is why breaking a protected extreme produces moves out of all proportion with cutting through a minor low.

The protected high is its exact mirror in a downtrend: the last high lower than the previous one, the one whose breach ends the fall, defended the same way by those who are short.

It is under that low, and above that high, that the external liquidity which counts sits. Not under any low: under the one the whole structure defends. The difference is enormous, and it shows in frequency.

Internal liquidity, by contrast, is swept constantly. The small lows and highs inside the move get cut through several times a session without announcing anything at all. That is the normal life of a market: price picks up what is lying around, fills medium-sized orders, and carries on. A beginner reading each of those passes as a signal manufactures ten signals a day, all of them false.

⛔ The consequence is the most concrete in the whole lesson, and it finally makes the earlier advice precise. A stop under a minor low will be taken, and often, because that spot is crossed daily by construction. A stop under the protected low will only go on the day structure genuinely gives way, which is exactly the day your reading is wrong and you want out.

“Put your stop a bit further away” is vague and arguable advice. “Put it behind the protected extreme” is checkable, can be pointed at on the chart, and gives the same answer to two people looking at the same thing. That is the usable version, and it is the one to keep.

⚠️ The price is real and worth stating: the protected low is often far. A stop placed there forces a smaller position, sometimes a much smaller one. That is not a flaw in the method, it is the honest cost of a stop that does not go off for nothing. If that cost makes the trade uninteresting, the risk lesson already gave the answer: the trade is too expensive for your account, and turning it down is correct.

Why the one on the other side has no stop like yours

Here is the point missing from almost every explanation of this subject, and it is very simply put.

You place a stop order. It waits, and if price touches it, it takes you out. That works because your position is small: when it closes, the market does not notice.

Whoever carries a thousand contracts cannot do that. Placing a stop for such a size would mean, on being triggered, sending an enormous order at once into a market already running against them. They would be filled worse and worse, at increasingly unfavourable prices, and their exit would cost far more than the planned loss. Their own stop would cost them more than the move that triggered it.

So they have no stop placed, and the consequence is the one worth keeping: they never get wicked out. The move that empties your account leaves them exactly where they were, because there is no order of theirs to trigger. On the same chart, at the same second, the two of you simply do not experience the same event.

They manage risk another way, and that is where the real difference sits. By choosing a size they can carry without being forced out. By hedging with another instrument, which caps the loss without placing an order. By scaling out in pieces, at their own pace, rather than all at once. And above all by choosing where they get in, so that price has little reason to come back for them.

⚠️ Do not conclude they work without a net. They have loss limits, often stricter than yours, and somebody watches them. Those limits simply do not take the form of an order lodged in advance at a price anyone can guess.

From which follows the point worth keeping from the whole lesson. They do not enter where it looks pretty, they enter where there are people on the other side. And the place with the most people on the other side is exactly where everyone else put their stop.

They need liquidity twice, not once

The part almost always forgotten, although it explains half of the moves that look absurd.

Getting in is one problem. Getting out is a second one, and it is exactly symmetrical. Whoever bought a thousand contracts into forced selling under a low now holds a thousand contracts. The day they want to sell them, they need a thousand buyers, and those are no more available than on the way in.

Where are buyers in quantity? Above a visible high, for exactly the same reason as below a low: that is where short sellers' stops sit, which trigger as buys, plus everyone waiting for a break to enter long. Two populations, one place.

That simple round trip makes a very common sequence readable: price drops to take the stops under the low, runs the whole range, takes the stops above the high, then falls back. Seen as a pattern it is incomprehensible. Seen as an entry then an exit, each served by its matching pool, it is almost mundane.

⚠️ The usual caution applies here more than anywhere: this reading explains a sequence after the fact very convincingly, and that is exactly what makes it dangerous. Not every visit to an extreme is a fill. A market can break a high because it is going up, quite simply. What you can build on solidly is not a forecast, it is a stop placement, and that is the subject of the next section.

The sweep, in three acts

Richard Wyckoff was already describing this sequence in the course he circulated in the early 1930s. The word upthrust, for a brief penetration above resistance that fails to hold, is genuinely his. The word spring, its mirror image below support, was settled later by the people who taught his method, and he is often given credit for it wrongly. The test he proposed, on the other hand, still works as written: a penetration that brings out no extra selling once the level is cleared has persuaded nobody to sell, it has merely collected what was already sitting there.

When the sequence happens it always looks the same, in three acts. Price grinds along the low, slowly, over several candles that resemble each other. Then it crosses, fast, on a single candle that covers in minutes what the previous ones took an hour to do. Then it leaves the other way, without coming back for anyone.

The second act is fast for an arithmetic reason. Triggered stops become market sell orders, and a sell order does not eat the offers, it eats the resting bids, working downwards one rung at a time. Every bid rung emptied ticks the price down a notch, which triggers the stops of the next rung. That is a cascade, and a cascade does not stop because it has gone far enough: it stops when there are no stops left to trigger, or when a buyer large enough puts themselves in the way.

The third act is the one that hurts. Once the band is empty, no sellers remain at that price, and whoever wanted to buy has their size. Price therefore rises with no effort, often past your entry, sometimes all the way to the target written on your plan. You were right about direction, and you are no longer in the trade.

⚠️ One expensive detail almost nobody mentions: during the second act, your stop does not execute at its price. It becomes a market order in the middle of a cascade, and it is filled lower, sometimes markedly lower. The real loss then exceeds the planned loss. This is not the only time that happens, the Sunday evening open and the seconds after a data release produce the same gap, but it is the only one of the three your placement has any grip on: a stop sitting on the busiest spot of the chart is also the one that slips the most.

What this reading does not explain

Here begins the honest part, the one most content on the subject carefully avoids. This is not a rule. Price does not go hunting stops every time, and the mechanism says why: somebody has to be holding an order too large for the book of the moment, and be holding it while the band is still there. Two conditions, and neither one is announced to you. An explanation of what happens often is in no way a prediction of what is about to happen, and the frequency on your instrument, at your hours, is not something to guess: it is something to count, and the exercise closing this lesson hands it to you in one evening.

Three outcomes are possible and nothing on the chart tells you which one you will get. Price sweeps then leaves: that is the case everybody talks about. Price sweeps and keeps going: your stop was right, your reading was wrong, and the stop did the exact job it exists for. Price never touches it: the move starts earlier, and whoever was waiting for a sweep to enter watches the train leave.

Your memory only keeps the first case. That is ordinary selection bias: an exit followed by a reversal is painful and memorable, an exit followed by a continuation is forgotten within days because it did you a favour. Counting by hand is the only way to know in what proportion the three outcomes show up on your instrument and at your hours. A written record remembers nothing, and that is precisely its virtue.

⛔ None of this is aimed at you, and none of it is coordinated. Your order is neither identified nor targeted, your position size interests nobody, and nobody holds a meeting to decide to go and fetch a low. What you are watching is a size constraint, public and mechanical: an order too large for the book can only fill where counterparty exists, and anyone can check that on an open depth ladder. The first lesson of this category sets out that constraint and its vocabulary, this one only draws the practical move from it. The idea that somebody is watching your screen is false, and above all very comfortable: it turns a placement error into an injustice, which excuses you from correcting it.

Where to place your stop

The practical consequence fits in one sentence: do not place your stop where everybody places theirs. One point under the exact low is the busiest spot on the chart, and the only spot you can be certain will be visited if price comes down that far.

The right instruction is not further away, which means nothing, but behind the zone. Your reading rests on a zone, a band of prices where you expect buyers, never on a line to the penny. The stop goes on the far side of that band, where price only travels if the zone has genuinely given way. An overshoot of three points invalidates nothing at all, a clean close on the other side invalidates everything.

A second reference completes the first, and this one is measurable. Put your stop beyond the maximum adverse excursion of your winning trades, a measure John Sweeney made known in the 1990s under exactly that name. The idea is direct: look at how far your winning trades went against you before leaving, take the worst, and place your stop behind it. A stop tighter than that number mechanically cuts a share of your winners, whatever the quality of your analysis.

⚠️ A wider stop is not a more permissive stop, it is a stop in the right place. It has to come with a smaller position, otherwise you have not moved an order, you have increased your risk. That is the subject of the next section, and it is the half of the lesson most people skip.

The trade-off is computed, not guessed

Widening the stop costs something, and the cost is exact. If you risk a fixed amount per trade, position size is that amount divided by the distance to the stop. Going from a twenty point distance to twenty-six points, thirty per cent further, shrinks the position by twenty-three per cent. Every winner of the month then brings in twenty-three per cent less.

On the other side, each pointless exit avoided hands you back a whole loss. Put both sides in the same unit, the risk of one trade, written R. Suppose your winners for the month produced fourteen times your risk: the wider stop shaves twenty-three per cent off that, so 3.2 R given up. If it avoids six pointless exits, it hands back 6 R. The balance is 2.8 R in favour of the wider stop.

That calculation has a threshold, and the threshold is the only number that matters. On the same assumptions, you need at least 3.2 pointless exits avoided in the month for the wider stop to pay. Below that, the tight stop remains the right choice, and widening it out of comfort would lose money. Nobody can hand you that threshold, because it depends on your instrument, your horizon and what your winners produce.

⛔ One mistake destroys the whole reasoning, and it is very common: widening the stop without shrinking the position. The distance changes, the amount at risk changes with it, and the entire month is then played on a risk base thirty per cent above the one that was planned. The trade-off only means something at constant risk. If you are not recalculating the size, do not touch the stop.

Two amplifiers: the mental stop and the thin session

The first amplifier is the mental stop, the order kept in your head instead of being placed. It is worth nothing, for a reason that has nothing to do with courage: it does not execute. A dropped connection, a meeting, a screen you do not look at for ten minutes, and the order that existed nowhere still exists nowhere.

The real reason is more awkward than the outage. The moment a mental stop ought to fire is precisely the moment your brain is least able to fire it. Daniel Kahneman and Amos Tversky formalised this in 1979 with prospect theory: facing a loss, a human being turns risk-seeking, while the same person is cautious facing an equivalent gain. Hersh Shefrin and Meir Statman drew from it in 1985 the disposition effect, the measured tendency to cut winners too early and hold losers too long.

A placed stop is therefore not an admission of weakness, it is a decision taken cold by the version of you that reasons, and imposed on the version of you that has just lost. No amount of character replaces that device. On a trading floor the loss limit is not held by the trader either: it is written into the system and watched by somebody else, precisely because nobody adjudicates their own case while losing. Your placed stop is the one-person version of that same device.

The second amplifier is the hour. The price move produced by an order depends on the depth of the book against it: Albert Kyle set out that relation in 1985, in a paper that founded the whole measurement of market impact. The practical consequence is brutal. The same quantity of stops triggered in a thin session, overnight, on a public holiday, or just before a release when everyone pulls their orders out of the book, moves price several times further than in a busy session.

That is why the most spectacular wicks appear at the hours when nothing is happening. The next day's chart shows a long spike and does not say it was built on a handful of transactions. If your schedule forces you to work in those windows, the stop band there is wider, your stop has to be wider too, and your position smaller in the same proportion.

Practising: count your pointless exits

Here is the exercise, and it produces a number worth more than the rest of this lesson put together. Take your losing trades from the past month, one by one. For each of them write down three things: the price of your stop, the extreme price reached against you, and what price did next, over the following two hours, or over the following twenty candles if you work on another timeframe.

Then sort each loser into one of three columns. Pointless exit: price came back above your entry and reached the target you had written down, without you. Justified exit: price kept going against you. Undetermined: neither of the two inside the window you observed. The third column is not a detail, it exists to stop you filing everything that annoyed you into the first one.

Count the first column. That number is your answer, and nobody else on earth can hand it to you. Then measure, on your winning trades from the same month, how many points price went against you before leaving, and keep the worst of them. That is the minimum distance your stop needed in order to cut none of your winners.

Finally apply the arithmetic of the previous section using those two numbers, which are yours. You get a verdict in numbers on a question many traders settle by intuition for years. A journal recording the maximum adverse excursion of every position logs it for you automatically, but a notebook and a pencil are plenty for one month: there are only a few dozen lines to write.

Run the count again the following month, once your stops have moved. It is the only check that exists. If the number of pointless exits does not fall while your positions have shrunk, your stop band was in the wrong place, and the fix is finding the real visible low rather than widening again. Two months of that record are worth more than two years of contradictory opinions on the right distance for a stop.