Chart patterns and what they describe

What a pattern is made of 

A chart pattern has two components and no third. The first is a set of swing highs and swing lows in a stated arrangement: how many of each, in what order, and at what levels relative to one another. The second is one or more straight lines drawn across those points, using the same construction the trend line lesson set out. Everything else attached to a pattern's name, what it is conventionally said to depict, how it is measured, when it is treated as void, sits on top of that construction rather than inside it. With those additions stripped away, what remains is a handful of turning points and a line.

The distinction from the candlestick arrangements in the previous lesson is one of scale and of raw material. A candlestick pattern is read from the bodies and wicks of one to three consecutive bars and needs no lines drawn on it at all. A chart pattern spans many bars, is largely indifferent to what any individual bar looked like, and cannot be stated without its lines. The two are routinely discussed in the same breath and they are different objects: one describes the shape of a few bars, the other the shape of several swings.

The arrangements below are the ones that appear in nearly every catalogue, and the descriptions of them are looser than their diagrams suggest. At about the same level, roughly parallel, a steep run, a short pause: each of those phrases hides a tolerance that is nowhere fixed. Two chartists applying different tolerances to one chart do not find the same set of patterns on it, and neither of them has made an error.

The reversal group 

The formations in this group share one feature of construction: a level is approached more than once, and a later approach does not carry beyond an earlier one. The group's name describes that shape in relation to the move that preceded it, and describes nothing else.

A double top is two swing highs at a comparable level with one swing low between them. The horizontal line drawn through that intervening low is called the neckline. The formation is conventionally described as complete once price closes below the neckline, and until that close the same three swing points describe an ordinary pullback inside an advance. A double bottom is the identical construction inverted: two swing lows at a comparable level, one swing high between them, a neckline drawn horizontally through that high, and completion conventionally taken as a close above it. A triple top or bottom extends the same description to three approaches and two intervening swings, with the neckline drawn through the lower of the two.

A head and shoulders is five swing points rather than three. Three swing highs, of which the middle one, called the head, is the highest, and the outer two, called the shoulders, sit at a comparable level below it, separated by two swing lows. The neckline is drawn through those two lows. Because the lows need not be level with one another, that line is frequently sloped rather than horizontal, which matters more than it first appears: a sloped neckline sits at a different price on the day it is reached than it did on the day it was drawn, so the completion level moves with time. The inverse head and shoulders is the same arrangement measured in lows, with the deepest low in the middle.

It is worth being exact about what a completed reversal formation establishes, because the name promises considerably more than the construction delivers. It establishes two things, both of them about prices that have already printed: that the arrangement occurred, and that price subsequently traded through a level derived from it. The word reversal in the label is a description of the shape's relationship to what came before, not a claim about what comes after.

The continuation group 

The formations in this group are drawn from two lines rather than one, and both lines are trend lines: one across successive swing highs, one beneath successive swing lows. What distinguishes the members of the group is the relationship between those two lines.

A triangle is the case where the two lines converge, and which line does the converging supplies the sub name. A symmetrical triangle has both lines sloping toward each other, the swing highs stepping lower and the swing lows stepping higher. An ascending triangle has a roughly horizontal line across the highs and a rising line beneath the lows. A descending triangle is its mirror, a roughly horizontal line beneath the lows and a falling line across the highs. Each line needs two swing points before it can be drawn at all, so a triangle is a four point object at minimum, and most catalogues want a further touch on at least one of the lines before they will use the word.

A flag is a short, shallow channel sloping against a steep and near vertical run immediately in front of it, that run being called the pole. A pennant is the same object with converging rather than parallel lines, which makes it a very small triangle sitting after a pole. A rectangle is two roughly horizontal lines, one across the highs and one beneath the lows, with price crossing between them: the range described in the trends and ranges lesson, drawn. A wedge is two lines converging while both slope the same way, which is the arrangement most often confused with a triangle, and the confusion is understandable because the difference is a matter of degree in the slopes rather than of kind.

Continuation, exactly like reversal, describes the arrangement's relationship to what preceded it. A triangle can resolve in either direction, and one that resolves against the preceding move is not usually recorded as a continuation that failed. It is renamed. That renaming is a convention of the vocabulary rather than an event in the market, and it is one reason catalogues of these shapes are difficult to compare with one another: two authors can agree completely about a chart and disagree about what was on it.

The measuring conventions 

Every formation in general use carries a measuring convention, an arithmetic rule that converts the pattern's own dimensions into a projected price level. The rule is nearly always the same in outline. The height of the formation is measured, and that height is then applied from the point at which price left the formation, in the direction price left it. The result is called a measured move.

Two properties of that arithmetic are worth stating plainly. It is entirely self referential: the projection comes from the formation's own dimensions and from nothing else, so a taller formation projects further for reasons of geometry rather than for anything observed about markets. And it describes a distance rather than an outcome. The calculation returns a level, and it returns the same level whether price goes on to trade there, stops short of it, or never approaches it at all.

The conventions also disagree with each other far more than their common outline suggests, and the disagreements are arithmetically material rather than cosmetic. For a head and shoulders, the height runs from the head to the neckline, but some authors measure vertically down to the neckline directly beneath the head while others measure to the neckline at the point where price crossed it, and on a sloped neckline those are two different heights and therefore two different projections. For a flag, the pole is measured and applied from the point price left the flag, though traditions differ over whether the pole starts at the last swing before the run or at the run's own first bar. For a triangle, one convention applies the height of the widest part of the formation while another draws a line parallel to the opposite boundary from the crossing point, and the two produce different levels on one drawing. A projected level quoted without the convention that produced it is incomplete in exactly the way a swing quoted without its confirmation setting is incomplete.

A pattern is only ever complete in hindsight 

The timing problem here follows directly from the earlier lessons. A swing point is confirmed only after a stated number of bars have closed on the far side of it, which means it is identified in arrears. A pattern is built from several swing points and at least one line drawn across them, so it inherits that delay once for every point it contains, and no line can be drawn until the points defining it exist.

Traced forwards rather than backwards, a double top arrives in this sequence. A swing high forms and is confirmed some bars later. A swing low forms and is confirmed some bars after that, at which point a horizontal line becomes drawable through it. Price approaches the earlier high again, and during that approach there is no second top, because nothing has yet stopped rising. A second swing high forms and is confirmed, and only now does the arrangement match the description at all. Then a close below the neckline, without which the whole structure is a pullback inside an advance that resumed. At every one of those moments the arrangement on the chart was compatible with more than one name, and the name that survived was selected by prices that had not printed yet.

When a formation stops matching its description 

Each formation has a condition under which the description no longer fits the data in front of it. These conventions are stated more consistently in the literature than most of what surrounds them, because they follow from the construction rather than from interpretation: the arrangement either still contains what the definition said it contained, or it does not.

• A double top is conventionally treated as abandoned once price trades above the higher of its two highs. The arrangement then contains a higher high, which is precisely the thing the description said had not happened. The double bottom's condition is the mirror of it.
• A head and shoulders is conventionally treated as abandoned once price returns above the head, and by some accounts earlier than that, once the right shoulder exceeds the head, since the middle swing high is then no longer the highest and the five points describe a different arrangement.
• A triangle has a finite life, because two converging lines meet. Traditions describe the crossing as conventionally occurring somewhere between half and three quarters of the distance to the apex, and treat price drifting into the apex without crossing either line as the formation dissolving rather than resolving. Where exactly that window sits is stated differently by different authors, and the apex itself is the only part of it that is not a matter of taste.
• A flag is defined by being small relative to the pole in front of it, so it is conventionally treated as abandoned once the counter move has retraced the whole of the run that preceded it. At that point the arrangement is not a pause in that run by its own definition.
• A rectangle is the one formation with no separate abandonment condition, because crossing either boundary is completion in one direction or the other and nothing is left over. What it has instead is an ambiguity: a boundary crossed and then recrossed leaves a wider rectangle rather than a completed one, and how far a crossing has to run before the wider reading is dropped is unfixed.

One clarification matters more than the list. Abandoned, in every line above, means the description has stopped matching the data. It says nothing about a position, an order or a decision, and none of these conventions carries an instruction of any kind. A formation ceasing to match its own definition is a fact about a drawing.

Where practitioners disagree 

The standing argument about chart patterns is whether they describe recurring market behaviour or supply a vocabulary loose enough to be applied to almost any series. The critical case is short: the descriptions carry unfixed tolerances, and a series with no structure in it whatsoever still contains arrangements that satisfy them, so finding a double top establishes that the definition was satisfiable and not much beyond that. The answering case is that the arrangements are shorthand for something real about how participants behave around levels that have already been tested, and that the geometry is a way of talking about that behaviour rather than a claim about shapes. The argument has run for decades in that form and is not close to a resolution.

Two narrower disputes sit underneath it, and both are visible on any annotated chart. The first is tolerance: how close two swing highs must be before they count as comparable, whether that closeness is measured as a percentage, as a multiple of recent range or by eye, and whether a formation found under one tolerance and absent under another was ever there. The second is volume. The codified tradition treats a volume condition as part of several of these definitions, most commonly that activity contracts as the formation develops and expands as price leaves it. Others treat volume as a description that often accompanies the shape rather than a requirement of it. The two traditions therefore do not identify the same formations on the same chart, which makes their catalogues quietly incompatible.

What none of these traditions supplies, and what this lesson does not supply either, is a figure for how often any arrangement is followed by any particular movement. Numbers of that kind depend entirely on the definitions and the tolerances used to find the patterns in the first place, which is the one thing every side of this argument is arguing about. A pattern is a description of prices that have already printed, and that is the whole of what it is.