What Elliott Wave and harmonic patterns claim

What the wave principle claims 

The wave principle states that price advances and declines in a repeating sequence: five legs in the direction of the larger movement, labelled with the numbers one to five, followed by three legs against it, labelled A, B and C. Of the numbered legs, the first, the third and the fifth run with the larger movement and the second and the fourth run against it. The claim is not that this sequence turns up from time to time. It is that the sequence is the structure of the series itself, so every stretch of every chart sits somewhere inside one, and locating it is the work the method asks for.

The second element is self similarity. Each leg is held to subdivide into the same eight part sequence at a smaller scale, and to be itself a single leg of the same sequence at a larger one. The tradition calls that scale degree, and degrees carry names running from spans of minutes to spans of decades. A labelled chart is therefore always a labelled chart at a chosen degree, and the same swing carries a different label at every degree above and below the one chosen. Nothing in the price series states which degree is in front of the reader.

Within that structure the tradition sets out three rules and treats them as inviolable. The second leg does not retrace the whole of the first. The third leg is not the shortest of the three legs running with the movement. The fourth leg does not enter the price territory covered by the first. Everything else, including the proportions most often quoted alongside a count, is stated as a guideline rather than a rule, which means a count departing from it is unusual rather than impermissible. That distinction carries most of the weight in this lesson, because those three rules are the only part of the system a chart is able to contradict.

What harmonic patterns claim 

A harmonic pattern is a five point structure, conventionally labelled X, A, B, C and D, joined by four legs. What separates it from the formations in the chart pattern lesson is that its definition is arithmetic rather than visual. Each leg has to stand in a stated proportion to another leg, and the proportions used are ratios derived from the Fibonacci sequence. A structure whose measurements match one published set of proportions is called a Gartley, a different set a bat, a third a butterfly, a fourth a crab. The names are arbitrary and the definitions are not, and that is the one respect in which this catalogue is more precisely specified than the classical one: a double top is a shape somebody recognises, while a harmonic structure is a set of measurements somebody checks.

The checking is where the precision leaks back out. No traded series produces exact ratios, so every published definition carries a tolerance, and the tolerance is stated differently by different authors: as a band of percentage points either side of the required ratio, as a zone bounded by two neighbouring ratios, or by eye. A structure belongs to a named family only relative to the tolerance it was tested against, and that tolerance is a setting rather than a fact about the market.

Two things are visible in that arithmetic. The measurement is exact and the membership is not, so a structure and its name are separated by a setting nobody has fixed. And the catalogue is open ended: as further families of proportions are named, the number of bands a given set of five points can fall into grows with it, so a set of points matching nothing under today's catalogue may match something under a larger one without a single price having changed.

Why both are difficult to test 

A claim can be checked only if some observation would count against it. That property is called falsifiability, and it is a low bar rather than a demanding one. A moving average either equals the mean of the closes it names or it does not. A swing high either satisfies the confirmation setting stated beside it or it does not. Both systems in this lesson are harder to hold to that bar than the rest of the module, by two different routes that arrive at the same place.

In the wave principle the three rules are stated over the labels, not over the prices. A chart cannot contradict a rule until labels have been attached to it, and the labels are chosen by the person reading it. When a rule is contradicted, the count is not thereby refuted. It is relabelled, either at a different degree, or as one of the named variants in which the contradicted rule does not apply. The stock of variants is large: corrections alone come as zigzags, flats, triangles and combinations of those, extensions lengthen a leg, truncations cut one short, and diagonals permit exactly the overlap the third rule otherwise forbids.

That block is not an argument against any individual count. It is a description of where the rules sit. A rule stated over labels can always be satisfied by different labels, provided the grammar holds a form in which the same prices are permitted, and this grammar holds a great many forms.

The harmonic route is different and finishes in the same place. The definitions are arithmetic, so a structure genuinely can fail one, as the worked block above shows. But the five points are chosen before they are measured, and nothing in any definition fixes which swings on a chart are to be taken as X, A, B, C and D. A chart of any length holds many candidate sets of five, and the sets that get measured are generally the ones that already resemble a named structure to the person measuring. Fitting a description to data after the data is known, using choices the description leaves free, is called retrofitting. It is not a failing peculiar to these two systems. It is the default behaviour of any system with enough free parameters, and both of these have a great many.

What follows from a method that can be relabelled 

One consequence is practical rather than philosophical, and it is the reason this lesson exists at all. A method that can be relabelled after the fact leaves no record of its own contradictions. The counts that were contradicted were replaced, so what survives on the annotated chart is the count that was never contradicted. The same holds for the harmonic structure that was measured and set aside before the one that fitted was found. The record left behind is a record of the survivors.

The wave tradition does supply the one thing that answers this, and it deserves the credit. A count carries an invalidation level, the price at which one of the three rules would be broken, and that level is stated in advance rather than afterwards. Where the invalidation is written down before the fact and the count is abandoned rather than relabelled once price reaches it, the framework produces a statement a chart can contradict, which is exactly the property the rest of this section says is missing. Where the response is an alternate count instead, it does not. That difference is a matter of practice rather than of theory, and descriptions of the wave principle routinely run the two together.

Why this appears once rather than as a track 

Every other lesson in this module ends by stating the conditions under which the thing it describes stops holding. A moving average lags by an amount set by its own window. An oscillator holds an extreme reading for as long as a strong move lasts. A trend line stops describing the data once price closes through it. That closing section is the module's organising principle, and it is the reason each method is taught alongside its own failure conditions rather than on its own terms.

A method whose contradiction can be repaired by relabelling has no such section available to write. There are no conditions to state, because the conditions are conditions on the labels, and the labels are free. Setting out a track of lessons on the wave catalogue would mean teaching that structure at length while never reaching the part every other lesson closes with, which would leave a reader with more detail about these systems than about any indicator in the module and less ability to say when either had been wrong.

The second reason is proportion. The material is genuinely large. The wave catalogue on its own would fill a module, and named harmonic families continue to be added. Length signals standing, whether or not it is meant to, and volume of detail is not a measure of how well a description has been tested. Recognition is the objective here, and recognition is short work: what is being claimed, what would have to be true for the claim to be checkable, and which count was kept once the others were dropped.

Where practitioners disagree 

The argument about the wave principle is not about whether markets have structure. Both sides accept that they do. It is about whether a grammar with this many permitted forms describes the structure or merely accommodates it. One tradition holds that the sequence is a description of collective behaviour, that the alternate counts are honest bookkeeping rather than evasion because each one carries its own invalidation level stated in advance, and that a framework offering several live readings at once is being candid about ambiguity the chart really contains. The other holds that a grammar able to accommodate any series describes none of them in particular, and that a rule which can be satisfied by relabelling is a convention about vocabulary rather than a constraint on data. The disagreement has run in that form for decades.

On harmonic structures the disagreement is narrower and more tractable, because the definitions are arithmetic. One side treats the ratios as a discipline: a requirement measured to three decimals cannot be argued into existence the way a shape can, and that is a real improvement on a catalogue of shapes with unfixed tolerances. The other side points at the two openings the arithmetic leaves, both of which the worked example above makes visible: the tolerance is a setting rather than a measurement, and the choice of which five swings to measure is unconstrained. Between them those two openings hand back much of the precision the ratios appear to supply.

No figure appears anywhere in this lesson for how often a count or a structure is followed by any particular movement, in either direction, and none appears elsewhere on this site. A number of that kind would depend on who did the labelling, at which degree, under which tolerance, and on which of the candidate five point sets were selected for measurement, and those are precisely the choices both systems leave to the reader. There is no honest number available to print, so none is printed, and the absence is the finding rather than a gap in the lesson.