Structure, levels, and volume

Support isn't a magic line. There's research using real order data explaining why price stalls in certain places and why it accelerates just after crossing them, and the mechanism is considerably more concrete than the name suggests.

How to define a trend unambiguously

A trend isn't a visual impression: it has a precise definition you can verify by checking two things, the highs and the lows.

The two definitions

Uptrend: the price marks progressively higher highs and progressively higher lows. Each rise reaches further than the last, and each pullback stops above where the previous one stopped.

Downtrend: the mirror image. Progressively lower highs and lower lows.

Notice both conditions are required. A price marking rising highs but progressively lower lows isn't in an uptrend: it's widening out, which is something else.

Where it starts breaking

Here's the useful part. Structure weakens before a trend reverses, and the warning is concrete.

In an uptrend, the first symptom isn't the price falling: it's that a pullback stops below where the previous one stopped. The rising low broke. The trend can keep climbing for a while after that, but one of its two conditions no longer holds.

It's an objective check with no interpretation: either the new low sits above the previous one or it doesn't.

What a range is, and why the answer depends on the timeframe

A range is a period where price oscillates between a ceiling and a floor without meeting either definition of a trend. How much time markets spend that way has no fixed figure — and not because nobody has measured it, but because it depends entirely on the scale you're looking from.

Why recognising it matters

Because tools that work in a trend fail in a range, and vice versa. A trend-following indicator inside a range generates contradictory signals one after another, and that's the cause of many bad runs for anyone trading fixed rules.

The figure that circulates, and why you won't read it here

You'll find on forums and in courses that markets trend 30% of the time and range 70%. You'll also find versions saying 15% and others saying 50%.

None of the three has a study behind it. They always appear as a rule of thumb, with no author, no market, and no period. They're exactly the kind of claim with a figure and no subject that this course teaches you to reject, and repeating it here just because it sounds concrete would be incoherent.

What has actually been measured

Work by Sara Safari and Christof Schmidhuber, of Zurich University of Applied Sciences and the University of Zurich, published in 2025, examined this across equities, interest rates, currencies, and commodities. They combined fourteen years of futures tick data, thirty years of daily prices, three hundred and thirty years of monthly prices, and yearly data reaching back to medieval times.

Their finding isn't a percentage, and that's why it's worth far more: markets sit in a trending regime on scales running from a few hours to a few years, and in a reversion regime on scales shorter and longer than that band.

In plain terms: look at a one-minute chart or a multi-decade one and what dominates is reversion to the mean. In the middle band — the one nearly everyone looks at — what dominates is continuation.

The practical consequence

The question "is the market ranging?" has no answer without stating on what timeframe. It's the same idea as the previous section, and here's why it isn't a technicality: a range on the hourly chart can be a perfectly orderly stretch of a weekly trend.

From which comes a concrete habit: before concluding something is going sideways, say which chart you saw it on. Without that qualifier, the statement means nothing.

The bias to guard against

Headlines cover markets surging or collapsing, because that's what constitutes news. A market that has spent four months oscillating without going anywhere appears nowhere.

That produces a distorted impression: it looks like the market is always doing something, when a considerable share of the time it isn't doing anything identifiable.

Recognising an asset is ranging, and accepting there's nothing to read there, is a legitimate conclusion and among the hardest to accept. Most analytical errors come from finding a story where there's only oscillation.

Why support and resistance genuinely form

Here's the part almost all material on this topic settles with "because lots of people watch them." There's a better explanation, and it comes from real data.

What support is and what resistance is

They are the same thing: a price level where enough orders pile up for the move to slow down when it gets there.

What changes is where the asset is trading.

If the price is above, that level is support. When it falls towards it, it meets buy orders, and the fall tends to slow.

If the price is below, the same level is resistance. When it rises towards it, it meets sell orders, and the rise tends to slow.

A stock's €50.00 is neither support nor resistance on its own. It is support if the stock trades at €54 and resistance if it trades at €46. The level is the same in both cases; what has changed is where the price is coming from.

One level, two rolesWhat decides whether it is support or resistance

€50.00 · the same level in both cases

The stock trades at €54

Support

The price is ABOVE. It falls towards the level, meets buyers and tends to bounce back up.

The stock trades at €46

Resistance

The price is BELOW. It rises towards the level, meets sellers and tends to bounce back down.

The €50.00 is neither support nor resistance on its own. What changes is not the level: it is where the price is coming from.

When the price goes through the level, the roles swap. A support that gives way ends up above the price, and from then on anyone looking at it will call it resistance.

That is not a theory about what will happen: it is arithmetic of position. The name depends on the side, and the side has just changed. How much confidence it deserves to expect it to also work in the new direction is another question, and it comes further down.

The study, and what data it used

Carol Osler published work in the Journal of Finance in 2003 examining something nobody had been able to see before: the individual stop-loss and take-profit orders of a currency dealing bank.

The data came from NatWest Markets, covered three currency pairs — dollar-yen, dollar-sterling, and euro-dollar — between August 1, 1999 and April 11, 2000, and included all customer orders and most of the bank's own.

What it found, and why it explains both things at once

The finding has two parts, and the combination is what's interesting.

Take-profit orders cluster strongly AT round numbers. These orders add supply when the price rises and demand when it falls, so they slow the move right at the level.

Stop-loss orders cluster just BEYOND the round number. These push in the same direction as the move, so they accelerate it once the level is cleared.

That asymmetry explains both classic technical analysis claims at once: that price tends to bounce at certain levels, and that it tends to accelerate when it breaks them. They aren't two phenomena: they're two consequences of the same order map.

How a support level formsOsler (2003) · Journal of Finance

SUPPORT · round number

Concentration of orders

Price falls to supportBouncesSupport gives wayFalls fast

Take-profit ordersStop orders

The same level holds many times and, when it finally gives, the move that follows is faster than usual.

Take-profit orders cluster AT the level and add demand, so they slow the fall. The stops of whoever bought sit just below, which is where everyone puts them: if the level gives way and the price reaches them, they trigger together and the fall accelerates.

Osler, C. (2003). Currency Orders and Exchange-Rate Dynamics. Journal of Finance.

Why round numbers

Because people pick them without coordinating. Traders favour prices ending in zero and five, and that preference produces order clustering at those specific points.

It isn't exclusive to currencies. Osler cites earlier work by Harris (1991) documenting the same in US equities: limit orders cluster at prices ending in 0 and 5, with stronger clusters at those ending in 0. And work by Niederhoffer and Osborne (1966) finding clustering at recent highs and lows too.

The two caveats that need stating

The data is from currencies, one specific bank, and an eight-month window. Extrapolating directly to any stock requires care, though the equity evidence points the same way.

And it explains the mechanism, not the profitability. That a concrete reason exists for the price behaving this way doesn't mean trading it is profitable after costs. Those are two separate questions and the second drags along every methodological problem from the previous session.

What is established, and it isn't nothing: there's a material, measurable reason behind levels, not just a shared belief.

How to tell a strong level from noise

Marking every high and low on a chart doesn't add information: it produces a chart where every price coincides with some line, and therefore where everything looks significant.

The problem with saturating a chart

Draw twenty levels on a screen and the price will constantly be touching one. Every touch will look like a reaction to an important level, and that sense of accuracy is mathematically inevitable with enough lines.

It's the same overfitting mechanism that closes this chapter, applied without a computer: the more hypotheses you draw, the more likely one fits by chance.

What makes a level strong, per what we know

From the previous section come concrete criteria rather than impressions:

That it's a round number. Order clustering at prices ending in 0 and 5 is documented in both currencies and equities.

That it coincides with a relevant recent high or low. There's evidence of order clustering at recent extremes too.

That several reasons point to the same place. A level where a round number, a previous high, and a long moving average coincide carries more weight than any of the three alone, because orders from three separate groups pile into the same zone.

A level is a zone, not a line

A widespread practical error: drawing support at €49.87 and treating €49.84 as a break.

The order clustering Osler described doesn't happen at an exact price: it happens around it. Plenty of people put their order at €50.00, some at €49.95, others at €50.10. The result is a band several cents wide — or several euros, depending on the asset — where activity concentrates, not a line.

Treating it as an exact line produces false precision: decisions hinging on a cent, triggering constantly on moves that mean nothing.

The sensible approach is drawing a zone whose thickness matches how much that asset moves. In a stock swinging €5 a day, a ten-cent zone makes no sense. That's exactly what the ATR (Average True Range) in the next session is for.

And what happens when a level breaks

You'll read everywhere that broken support becomes resistance, and vice versa. That claim deserves the same criterion as the rest of the chapter.

What can be said with grounding: the order map that existed at that level changes after a break. The take-profit orders that were slowing the price have executed, and so have the stop orders on the other side. The level no longer has what it had.

What can't be asserted: that the level will now work in the opposite direction with any measurable reliability. It's a reasonable hypothesis about a real mechanism, not a demonstrated result.

The distinction matters because it separates "something was here and now isn't" — descriptive and verifiable — from "this will stop the price in the opposite direction" — a prediction without citable support.

What can't be asserted, and not for want of looking

How many touches make a level "strong," or which combination of factors best predicts whether it holds or breaks, has no quantified answer we could cite honestly.

Volatly went looking for that answer across its own universe of assets and closed the study in July 2026. The question was whether a more conspicuous level — more touches behind it, respected for longer, more obvious on the chart — holds better than an unremarkable one. The answer was no. No way of measuring how much a level stands out separated the ones that stalled the price from the ones that didn't.

Seen through the mechanism this chapter describes, that fits. A level isn't a wall: it's a build-up of orders in a zone. And a build-up of orders gets used up. Given enough volume in the other direction — earnings nobody saw coming, a large fund that needs out — those orders fill one after another until there are none left, and the price carries on. However many times it held before makes no difference. What was holding were the orders sitting there, and orders run out.

Which is why neither this course nor Volatly will tell you a level is going to stop the price. What can be said is where it sits, how many times it has been touched, and what happened after each touch. That's checkable description, and it's a different thing from a prediction.

Publishing an invented number about how many touches make support strong would be exactly what this course teaches you to distrust, done by the course itself.

When volume confirms and when it contradicts

Volume measures how many people participated in a move. It doesn't say where the price is going: it says how much conviction was behind what already happened.

The basic reading

A move on volume well above normal indicates many people participated. A move on low volume indicates few did.

Applied to a level break: if the price clears resistance on high volume, plenty of people are behind that move. If it clears on quiet volume, there's more reason to suspect a false breakout — the price crosses briefly and returns, because there wasn't enough backing to sustain it.

Divergence, which is the most useful part

Here's the signal that contributes most in this section.

If the price keeps marking higher highs but the volume accompanying each new high keeps shrinking, fewer and fewer people are pushing that move. The price doesn't reflect it yet, but participation is running out.

It isn't a guaranteed reversal signal. It's a warning that the force behind the move is diminishing, which is different and more honest information.

The limits worth knowing

Volume confirms better than it predicts. It serves to ask whether enough people were behind what just happened, not to guess what comes next.

And not all volumes are comparable. As we saw in the first chapter, activity is U-shaped through the day and spikes around scheduled events. High volume on an earnings day says nothing special: it says there were earnings.

Comparing today's volume against that asset's average volume, rather than against an absolute figure, is the only way the reading means anything.

What to remember

• A trend requires two conditions at once: rising highs and rising lows. One failing is the first warning.
• Osler documented with real orders that take-profit orders cluster at round numbers and stop orders just beyond: that explains both bounces and accelerations.
• How many touches make a level strong has no citable quantified answer, and saying so is more honest than inventing one.
• Volume confirms what already happened and only informs compared against that asset's normal volume.

Related: what actually happens when you hit buy · order types and how they execute · why price history changes after a split