Market Structure: Trends, Ranges and Transitions

“The trend is up” is the most frequently asserted sentence in technical analysis and one of the least often defined. This lesson gives it a definition, applies the definition mechanically, and then does the thing that most treatments skip: shows you what the definition is resting on, so that when someone disagrees with your classification you can find out whether you disagree about the market or merely about the parameters.

By the end you will be able to classify any chart into one of three states using a stated rule, state how late that classification necessarily is, and use the regime vocabulary that Parts 12, 27 and 35 depend on.

The four comparisons

Take the confirmed swing points from the previous lesson and line them up in time. The structure vocabulary is built from exactly four comparisons between consecutive swings of the same kind:

Comparison	Meaning
Higher high	The latest confirmed swing high is above the one before it
Higher low	The latest confirmed swing low is above the one before it
Lower high	The latest confirmed swing high is below the one before it
Lower low	The latest confirmed swing low is below the one before it

Combining them gives the familiar labels. Higher highs and higher lows together are what most people mean by an uptrend. Lower highs and lower lows together are a downtrend. And the combinations that are left over — higher highs with lower lows (expanding), lower highs with higher lows (contracting), and every case where one comparison is a tie — are not trends at all.

That last group is the interesting one, because it is where a great deal of market time actually sits, and because most chart commentary quietly reassigns it to whichever trend the commentator already believed in.

Drawn, the four comparisons stop being vocabulary and become two things you can measure with a ruler:

Higher highs and higher lows

• Close above open
• Close below open
• Swing high
• Swing low

Show the numbers behind this chart

Bar	Open	High	Low	Close
1	100	101	99	100
2	100	102	100	102
3	102	104	102	104
4	104	107	104	106
5	106	106	105	105
6	105	105	103	103
7	103	103	101	101
8	101	103	101	103
9	103	106	103	106
10	106	109	106	109
11	109	112	109	111
12	111	111	109	109
13	109	109	107	107
14	107	107	105	105
15	105	108	105	108
16	108	111	108	111
17	111	114	111	114
18	114	117	114	116

Three swing highs, each above the one before it, and two swing lows, the second above the first. That is the entire content of the claim 'this is an uptrend' - two pairs of numbers compared. The data in this chart is invented for the illustration. It is not market data and nothing should be inferred from it.

What the label asserts, and what it does not

Read the definition again and notice how modest it is. “Higher highs and higher lows” is a statement about two pairs of numbers taken from the past. It asserts that the most recent confirmed swing high exceeded the previous one and the most recent confirmed swing low exceeded the previous one. That is all it says.

It does not say the next swing will also be higher. It does not say a rise is more likely than a fall from here. It does not say anything about how long the condition will persist. Those are all separate claims — some of them testable, none of them contained in the classification itself.

?Reality check

The popular claim is: “the trend is your friend — an established uptrend is more likely to continue than to reverse.” Notice that this is not a definition but an empirical hypothesis, and that it becomes measurable the moment you fix a swing definition. Define uptrend as above, define “continue” as making one more higher high before making a lower low, and count. Part 12 builds exactly this kind of measurement, and Part 28 does it with a full backtest. The honest position until then is that you have a vocabulary, not evidence.

Ranges are a real answer, not a failure to classify

A range is what you have when the swing highs and swing lows are not consistently moving in the same direction. It deserves a name of its own for a practical reason: almost every technique that works on directional structure behaves differently inside a range, and many behave badly. Moving-average crossovers whipsaw. Breakout entries fail repeatedly at the same level. Momentum oscillators oscillate, which is when they look most convincing and mean least.

There is no bright line between “trend” and “range”, and pretending otherwise is where a lot of self-deception starts. What you can do is refuse to force a third answer into one of the first two. The formula in this lesson has an explicit third label, Unclassified, and on most instruments it is on screen more often than either directional label. That is not the formula being unhelpful. That is the chart.

Structure breaks

A structure break is the bar on which the pattern of comparisons changes: the first lower low after a run of higher lows, or the first higher high after a run of lower highs. It is the transition, and it is the thing traders most want to catch early.

Two cautions travel with it.

First, it is an event in the precise sense the course uses that word, whereas the classification either side of it is a state. The distinction sounds pedantic here and becomes the difference between a working formula and a broken one in Part 9. A state is true over a span of bars; an event is the bar on which something became true. Asking “is the structure up?” and asking “did the structure just turn up?” are different questions requiring different code.

Here is the same series continued. Nothing about the first eighteen bars has changed; the chart is longer, and one comparison has flipped.

The same series, eight bars later

• Close above open
• Close below open
• Swing high
• Swing low
• Lower low - the structure break

Bar 23 trades to 101.50, below the prior swing low of 104.50. That single bar is the event; every bar before it was the state. Part 9 turns that distinction into two different lines of AFL. The data in this chart is invented for the illustration. It is not market data and nothing should be inferred from it.

Second, a structure break is only as timely as the swings it is made of. If your swing strength is 5, the lower low that constitutes the break was confirmed 5 bars after it happened — and the “break” you are reacting to is therefore at least 5 bars old before you ever see it.

What an “uptrend” claim actually rests on

The layers under one word

each layer depends on the one below

1. "The trend is up"What gets said out loud
2. Comparison of the last two of each swingHigher high AND higher lowno free parameters
3. Set of confirmed swing pointsOne Boolean per barlagged by n bars
4. Swing definition and its strength nChosen by you, not by the marketthe free parameter
5. Price series at one intervalDaily, weekly, 5-minute…also chosen by you
6. The trades that actually happenedThe only layer that is not a choice

Two people can disagree about the top layer while agreeing about every trade that took place. The disagreement lives in the two layers marked as choices.

This diagram is the argument of the lesson. When someone says a chart is in an uptrend and you think it is not, the productive question is never “who is right?” It is “what swing strength and what interval are you using?” More often than not the disagreement dissolves, and when it does not, you have learned something specific about how sensitive the conclusion is.

The regime vocabulary this course uses

From here on, these words carry fixed meanings. They are used consistently in every later part, and the later parts will not redefine them.

Term	Meaning in this course
Market regime	A classification of overall market conditions, used as a gate on whether to act at all
State	A condition true over a span of bars, such as higher highs and higher lows
Event	The single bar on which something became true, such as a structure break
Setup	The conditions that make an instrument a candidate
Trigger	The specific event that turns a candidate into an entry
Timeframe	The conceptual horizon, such as weekly context with daily setups
Interval	The concrete bar size — inDaily, five-minute — as AmiBroker means it

The distinction between timeframe and interval matters more than it looks. “I trade the daily timeframe” is a statement about horizon; inDaily is a setting. Part 14 spends a whole part on the machinery for using two intervals at once without accidentally reading the future.

The formula

Goal

Take the confirmed swing points, keep the last two of each kind, apply the four comparisons, and display the resulting label continuously — in words, and as a background ribbon — together with the four swing prices the label was derived from. The purpose is not to have a trend indicator. It is to make your visual classification falsifiable: when the ribbon disagrees with your eye, one of you is applying a different rule, and now you can find out which.

Complete formula

Complete runnable AFL

market-structure.afl

// market-structure.afl
// Part 4 - Market Structure: Trends, Ranges and Transitions
//
// Classifies the chart into one of three labels using the confirmed swing
// points from the previous lesson:
//
//   higher highs and higher lows   -> "up structure"
//   lower highs and lower lows     -> "down structure"
//   anything else                  -> "unclassified"
//
// The classification is a definition applied consistently. It describes what
// the recent swing points did. It says nothing about what comes next.
//
// Assumptions:
//   - Swing strength is the only free parameter. Change it and the labels
//     change; comparing two settings is the honest way to use this formula.
//   - Every label is late by at least Strength bars, because a swing point
//     cannot be confirmed until Strength bars have printed after it.
//   - "Unclassified" is a real answer and, on most instruments, a common one.

_SECTION_BEGIN("Market structure");

Strength = Param( "Swing strength (bars either side)", 5, 1, 25, 1 );
Window   = 2 * Strength + 1;

BarNumber = Cum( 1 );
HasWindow = BarNumber >= Window;

CandidateHigh = Ref( High, -Strength );
CandidateLow  = Ref( Low,  -Strength );

SwingHigh = HasWindow AND CandidateHigh == HHV( High, Window );
SwingLow  = HasWindow AND CandidateLow  == LLV( Low,  Window );

// ValueWhen's third argument selects the n-th most recent occurrence:
// 1 is the latest, 2 is the one before it. Two of each is the minimum needed
// to say whether highs and lows are rising or falling.
HighNow  = ValueWhen( SwingHigh, CandidateHigh, 1 );
HighPrev = ValueWhen( SwingHigh, CandidateHigh, 2 );
LowNow   = ValueWhen( SwingLow,  CandidateLow,  1 );
LowPrev  = ValueWhen( SwingLow,  CandidateLow,  2 );

HigherHigh = HighNow > HighPrev;
HigherLow  = LowNow  > LowPrev;
LowerHigh  = HighNow < HighPrev;
LowerLow   = LowNow  < LowPrev;

UpStructure   = HigherHigh AND HigherLow;
DownStructure = LowerHigh AND LowerLow;

Plot( Close, "Price", colorDefault, styleCandle );

// The ribbon is a convenience, not the message. Colour is never the only way to
// read the classification here: the title states it in words as well, which is
// also what makes the chart usable by a reader who cannot distinguish the hues.
RibbonColour = IIf( UpStructure, colorPaleGreen,
               IIf( DownStructure, colorRose, colorLightGrey ) );

Plot( 2, "Structure", RibbonColour,
      styleOwnScale | styleArea | styleNoLabel | styleNoTitle, -0.5, 100 );

StructureText = WriteIf( UpStructure, "Higher highs and higher lows",
                WriteIf( DownStructure, "Lower highs and lower lows",
                         "Unclassified - the swings disagree" ) );

_N( Title =
    Name() + " - " + Interval( 2 ) +
    StrFormat( " - structure at swing strength %g\n", Strength ) +
    StructureText + "\n" +
    StrFormat( "Swing highs: %g then %g\n", HighPrev, HighNow ) +
    StrFormat( "Swing lows:  %g then %g\n", LowPrev, LowNow ) +
    StrFormat( "This label could not have been known until %g bars "
             + "after the newer swing formed.", Strength ) );

_SECTION_END();

Download market-structure.afl75 lines

How it works

The first half is the swing detector from the previous lesson, unchanged. The second half is new and short.

ValueWhen() takes a condition, an array, and an occurrence number. Called with 1 it gives the value the array had at the most recent bar where the condition was true; called with 2, the occurrence before that. Four calls therefore give the last two confirmed swing highs and the last two confirmed swing lows, each carried forward across every bar until a newer one replaces it — a step-shaped series, which is exactly what a classification needs.

The four comparisons are then plain array comparisons producing one Boolean per bar, and the two labels are AND combinations of them. Anything that is neither is left unlabelled rather than being assigned to the nearest available trend.

The ribbon uses the documented idiom from AmiBroker’s own styles tutorial: plot a constant with styleOwnScale | styleArea, where the constant is the ribbon height as a percentage of the pane, and pass a colour array rather than a colour, so the colour changes bar by bar. The minimum and maximum arguments are only used by styleOwnScale plots, which is why they appear here and nowhere else in the part.

!Important

The label is also written in words in the title, and that is not decoration. A classification encoded only as a background colour is unreadable to a reader with a colour-vision deficiency, unusable in a printed handout, and impossible to quote in a written brief. Any time you colour something in this course, ask what the reader who cannot see the colour is supposed to do.

Key functions

• ValueWhen( expression, array, n ) — condition first, then the array, then which occurrence. The official page notes that n also accepts zero and negative values, which reference future occurrences; that is a look-ahead trap, and this formula uses only 1 and 2.
• WriteIf( expression, "true text", "false text" ) returns one of two strings. Despite the name it writes nothing; it is a conditional text chooser, and nesting two of them gives the three-way label.
• Plot( array, name, colour, style, minvalue, maxvalue ) — the fifth and sixth arguments are documented as being used by styleOwnScale plots, which is what makes the ribbon sit at a fixed height regardless of the price scale.
• IIf( condition, valueIfTrue, valueIfFalse ) selects element-wise, per bar. Nested here to build the three-colour array.

Expected result

=Expected result

A candlestick chart with a coloured band behind it: pale green while the last two swing highs and last two swing lows are both rising, pale rose while both are falling, and light grey otherwise. The title states the same thing in words and lists the four swing prices, so you can check the label against its inputs at any bar. On most instruments the grey band covers a large share of the chart.

Test it

Do not test this by looking at whether it “gets the trends right” — that judgement is exactly the subjective thing the formula exists to replace. Test it against its own definition:

1. Select a bar where the label is “higher highs and higher lows”. Read the four swing prices from the title. Confirm by arithmetic that the newer high exceeds the older and the newer low exceeds the older.
2. Find a bar where the label is “unclassified” and read the four prices. Confirm that at least one of the two comparisons fails.
3. Find the exact bar where the label changes. Confirm that a new swing was confirmed on that bar — the label can only change when new information about a swing arrives.

Common errors

• The ribbon is grey for the entire chart. Almost always too few confirmed swings: either the strength is too high for the loaded history, or you are looking at fewer bars than 4 × Strength and the second occurrence of a swing does not exist yet.
• The label changes on a bar with no new swing. It cannot, by construction. If it appears to, you are reading the ribbon at a different bar from the title — the title reports the selected bar, and with no selection it reports the last visible bar.
• The ribbon covers the candles. Plot order inside a Z-order layer runs in reverse call order by default, so the last plot in the code is drawn furthest back. If you reorder the plots and the ribbon comes forward, that is why.
• ValueWhen arguments swapped. ValueWhen( CandidateHigh, SwingHigh ) compiles perfectly and returns nonsense. The condition always comes first.

Extension

Add a second classification using a different swing strength — 3 alongside 10 — and plot both ribbons in separate thin bands. The bars where the two disagree are the honest answer to “is this a trend?”, which is: it depends on the scale you are asking about, and here is exactly where that matters.

Run it twice before you believe it

Whatever the ribbon says at strength 5, run it again at strength 2 and at strength 12 on the same chart, and note the three answers.

If all three agree, you have a classification that is robust to the parameter, and you can quote it with some confidence. If they disagree, you have learned that the label is an artefact of a setting rather than a description of the instrument, and the correct thing to say out loud is “unclear at this scale” rather than picking the setting whose answer you preferred.

This two-line habit — vary the parameter, report the spread of answers — is the cheapest robustness check in the whole course, and it generalises directly to optimisation in Part 31, where the same idea appears in a much more expensive form.

What changed

You now have a three-way classification built from four comparisons between confirmed swing points, and you know precisely what it rests on: an interval you chose, a swing strength you chose, and arithmetic that has no opinions. You know it is late by at least the swing strength, that the third category is not a failure, and that the way to settle a disagreement about a chart is to compare parameters rather than convictions.

The lab that follows turns all of this into a procedure with an order and an output, so that reading a chart becomes something you do the same way twice.

Check your understanding

Question 1. The last two confirmed swing highs are 118 then 124. The last two confirmed swing lows are 109 then 104. How does the classification in this lesson label this?

• Higher highs and higher lows
• Lower highs and lower lows
• Unclassified
• It cannot be determined without the closing prices

Show the answer and why

Answer: Unclassified

The highs are rising and the lows are falling, so neither directional definition is satisfied. This expanding pattern is a genuine third case, and forcing it into "uptrend" because the highs rose is exactly the reassignment the lesson warns about.

Question 2. Your swing strength is 8. A structure break occurs on bar 200. What is the earliest bar on which your formula can report it?

• Bar 192
• Bar 200
• Bar 208
• Bar 216

Show the answer and why

Answer: Bar 208

The break is defined by a swing point, and a swing point of strength 8 is not confirmed until 8 bars have printed after it. The classification inherits the whole of the swing definition's lag; nothing downstream can be more timely than its inputs.

Question 3. Two analysts classify the same daily chart differently. What is the most productive first question?

• Which of them has more experience?
• What swing strength and interval is each of them using?
• Which one used candlesticks?
• What does the moving average say?

Show the answer and why

Answer: What swing strength and interval is each of them using?

A structure label is derived from a swing definition applied to an interval. Both are chosen, not observed, so two correct applications of different parameters produce different labels from identical data. Comparing parameters usually dissolves the disagreement or locates it precisely.

Question 4. Which of these are statements the classification actually makes? Select all that apply.

• The most recent confirmed swing high is above the one before it
• The most recent confirmed swing low is above the one before it
• The next swing high will be higher still
• A rise from here is more likely than a fall

Show the answer and why

Answer: The most recent confirmed swing high is above the one before it, The most recent confirmed swing low is above the one before it

The label is a comparison of four past numbers and contains no claim about what follows. The last two options are separate empirical hypotheses which may be true, false or conditional, and which require measurement rather than definition.

Question 5. Why does the formula print the classification as text in the title as well as colouring a ribbon?

• Because the ribbon colours are unreliable in AmiBroker
• Because meaning encoded only in colour is unreadable to some readers and unquotable in writing
• Because the title updates faster than the ribbon
• Because AFL cannot colour a ribbon conditionally

Show the answer and why

Answer: Because meaning encoded only in colour is unreadable to some readers and unquotable in writing

Colour-only encoding fails for readers with a colour-vision deficiency, in print, and in any written brief. AFL colours ribbons conditionally without difficulty — the text is there for accessibility and for the chart brief you write in the next lesson.

Check my answersClear