Elliott wave theory: the rules explained

Elliott wave theory: the definition

The theory carries the name of Ralph Nelson Elliott, who published The Wave Principle in New York in 1938, a 50-page text reproduced from a typewritten copy, according to its record on Open Library. StockCharts' ChartSchool, in its introduction to Elliott Wave Theory, writes that the theory was developed by Elliott in the 1930s and popularised by Robert Prechter in the 1970s. The 1938 text is not freely available, and the reference textbook, Elliott Wave Principle by A. J. Frost and Robert Prechter, can be read online on Elliott Wave International's website after registering: what this lesson attributes to the theory comes from secondary sources named each time, and no sentence is attributed to Elliott himself. Pages consulted in October 2026.

StockCharts' definition fits in one sentence: a method of market analysis based on the idea that the market forms the same types of patterns on a smaller timeframe, called a lesser degree, as it does on a longer one, called a higher degree. The introduction to the Wave Principle published by Elliott Wave International, the company founded by Robert Prechter, sets out the basic tenet: an impulsive wave is composed of five subwaves and moves in the same direction as the trend of the next larger size; a corrective wave is divided into three subwaves and moves against that trend.

Impulse waves are numbered 1 to 5, corrective waves are lettered A, B, C. Waves 1, 3 and 5 move with the trend, StockCharts calls them 'actionary'; waves 2 and 4 move against it. The same numbers and the same letters are used in every language, which is why the diagrams in this lesson carry nothing else.

Elliott, born in 1871 according to Elliott Wave International's page, is a contemporary of Richard Wyckoff, and the two men read the same American stock market with two different grids: Wyckoff adds volume to price, Elliott looks only at the shape of price. The lesson on the Wyckoff method and VSA details the first; the lesson what is technical analysis places both in the history of the discipline.

Five waves, three waves: the structure and the degrees

The full cycle has eight waves: five with the trend, numbered 1 to 5, then three against it, lettered A, B, C. In the diagram, price rises from 100 to 128 in five waves, then gives back part of the advance in three. In a downtrend, everything is reversed: five waves down, three up.

What sets the theory apart from a mere division is the nesting. Roy Batchelor and Richard Ramyar sum it up this way in 'Magic numbers in the Dow': Elliott's basic idea is that the market typically rises in five phases and then falls in three, that this pattern is self-similar and can be seen at all data frequencies, so that within each long-term wave there are five rising and three falling phases, and so on. An Elliott wave might therefore be observed in a century of stock market history as in a chart of five-minute bars. Each level of nesting is called a degree.

Hence the notation of subwaves. A five-wave impulse breaks down as 5-3-5-3-5: its waves 1, 3 and 5 are themselves five-wave impulses, its waves 2 and 4 three-wave corrections. An A-B-C correction breaks down according to its form, 5-3-5 for a zigzag, 3-3-5 for a flat, as the section on corrections details. StockCharts notes that waves 1, 3 and 5 are usually motive waves themselves, since they move in the direction of the trend of one larger degree.

Counting waves therefore means saying where the market stands in that cycle: in wave 3, usually the strongest, in wave 5, the last, or in a correction. That is the method's promise, and it is also what makes it hard to verify, as the end of this lesson shows.

The three rules of impulse waves

The theory distinguishes rules, which allow no exception, from guidelines, which describe a frequent tendency. An article by the editorial staff of Elliott Wave International says it in two sentences: rules tell you what you can't do, guidelines tell you what is probable. It states the three rules of an impulse as follows: wave 2 can never retrace more than 100% of wave 1; wave 3 can never be the shortest of waves 1, 3 and 5; wave 4 can never end in the price territory of wave 1.

StockCharts states the same rules with one difference: its introduction page writes that wave 4 always retraces less than 100% of wave 3, and its page on identifying Elliott wave patterns adds that wave 4 can never overlap wave 1. The two wordings complement each other, they do not contradict each other. What they share matters more: a rule never says what the market will do, it only says that a count is wrong if the rule is broken.

The exception has a name: the diagonal. StockCharts describes the ending diagonal, in the wave 5 position, and the leading diagonal, in the wave 1 position, as structures where waves 2 and 4 overlap and form a wedge with converging boundary lines. In a diagonal, the same page writes, wave 3 still cannot be the shortest, and each actionary subwave never fully retraces the previous one. What looks on a chart like the chartists' wedge is given a name here and a place in the count.

Guidelines are of another order. StockCharts' page on the guidelines warns that a guideline is not a hard and fast rule that can't be broken: it is a tendency, something that happens so often that it can almost qualify as a rule, but at times doesn't work as expected. It lists several: alternation, whereby if wave 2 is a sharp correction, wave 4 will be a sideways correction; equality, whereby when wave 3 is the extended wave, wave 5 will approximately equal wave 1 in price; the depth of corrections, which often bring the market back to the territory of the previous wave 4 of lesser degree; and channeling, whose parallel lines often mark the boundaries of an impulse.

Corrections: zigzag, flat, expanded flat, triangle

A correction moves against the trend of the higher degree and is written in letters. StockCharts' page on identifying patterns describes three simple families and one combined family, and the diagram below draws them side by side.

The zigzag is a three-wave A-B-C correction whose subwaves form a 5-3-5 structure: A and C are five-wave impulses, B a three-wave correction. It is the sharp correction, the one that goes far. Two or three zigzags can follow one another as a double or triple zigzag, linked by corrective waves.

The flat is a three-wave correction whose subwaves form a 3-3-5 structure: A and B are corrections, only C is an impulse. It moves sideways, and StockCharts notes that it frequently appears in the wave 4 position. The regular flat ends without extending beyond its starting point. The running flat, rare and found in strong trends, has a wave B that ends beyond the start of A but a wave C that fails to reach the start of A.

The triangle has five subwaves in a 3-3-3-3-3 structure, lettered A-B-C-D-E, between two converging lines, or diverging ones for the expanding triangle; symmetrical, ascending or descending depending on the slope of the lines. StockCharts writes that it always appears in the position prior to the final move, that is as wave 4 or as wave B. It is a relative of the chartists' triangle pattern, with one difference: here the number of legs is fixed at five and the triangle has a place in the count.

Combinations, finally, string several simple corrections together, written W-X-Y for a double and W-X-Y-X-Z for a triple. This wealth of forms is also the method's weak point: almost any sequence of three moves against the trend finds a name in the catalogue, and the section on limits comes back to it.

The expanded flat Elliott wave: B beyond the start of A, C beyond its end

The expanded flat is the form of flat that traders look for most, because it traps twice. StockCharts defines it by two overshoots: wave B ends beyond the start of wave A, and wave C ends beyond the start of wave B, that is beyond the end of A. In an uptrend, B therefore makes a new high above the top of wave 5, then C falls below the low of A.

That is the trap. B's new high looks like the trend resuming; it is in fact the middle of a correction, and C then takes out the buyers who came in on that high. What the theory describes here, the article on the break and retest describes from the other side: a breakout that does not hold. The orange, in both diagrams of this lesson, marks this form.

Secondary sources give it proportions. Eugenio D'Angelo and Giulio Grimaldi, in a 2017 article published in International Business Research, whose theoretical diagrams are adapted from Frost and Prechter, write that in a regular flat waves A, B and C are typically approximately equal, and that in an expanded flat wave C is often 1.618 times the length of A. These are guidelines, not rules: the definition of the expanded flat rests on the two overshoots, not on a ratio.

A worked example: five waves from 100 to 128, then an expanded flat

The prices are invented and the first diagram of this lesson draws them to scale. Starting point 0 at 100. Wave 1: from 100 to 110, 10 points. Wave 2: from 110 to 104, it gives back 6 points, that is 6 ÷ 10 = 60% of wave 1. Rule 1 holds: 104 stays above 100. Wave 3: from 104 to 124, 20 points, twice wave 1. Wave 4: from 124 to 116, it gives back 8 points, that is 8 ÷ 20 = 40% of wave 3, and 116 stays above 110, the top of wave 1: rule 3 holds. Wave 5: from 116 to 128, 12 points. The three actionary waves measure 10, 20 and 12: wave 3 is not the shortest, rule 2 holds.

The guidelines, on these numbers. Wave 2 gives back 60% and wave 4 gives back 40%: the first is deeper than the second, which is the most common form of alternation. Wave 3 is the extended wave, twice wave 1, and wave 5 runs 12 against 10 for wave 1: close to equality, not equal. None of these ratios is exactly a Fibonacci ratio, and that is deliberate: 60% is close to 0.618, 40% close to 0.382, and the glossary entry on the Fibonacci retracement explains why almost any pullback seems, after the fact, to stop near a level.

The expanded flat. After the top of wave 5 at 128, A falls to 120, 8 points. B rises to 130: 10 points, 1.25 times A, and 130 exceeds 128, the start of A, first overshoot. C drops to 117: 13 points, 1.625 times A, and 117 goes below 120, the end of A, second overshoot. StockCharts' two conditions are met. The ratio of 1.625 is close to the 1.618 that D'Angelo and Grimaldi give for wave C of an expanded flat, without being equal to it.

What the correction gave back. The advance ran from 100 to 128, 28 points. The correction stops at 117, so it gave back 128 − 117 = 11 points, that is 11 ÷ 28 = 39% of the advance. And 117 lies between 116 and 124, the territory of wave 4: that is the guideline on the depth of corrections. On this finished chart, everything reads. Live, at the moment B prints 130, nothing distinguishes that high from a wave 1 of the higher degree, and that is the whole subject of the next section.

The trade the theory suggests, and its risk. A trader who counts an expanded flat sells after the top of B, at 129 for example, with a stop above 130, say 131, that is 2 points of risk, and aims for the territory of wave 4, 124 at the nearest: 5 points, 2.5 times the risk. If the count is wrong and 130 was a wave 1 or 3 of the higher degree, the stop is hit at 131 and the trade costs 1R. The article on how to backtest a trading strategy says how to count those two outcomes on your market before risking anything on it.

What published research says, and does not say

The ratios, tested on the Dow. Roy Batchelor and Richard Ramyar, of Cass Business School, examined in 2006 whether ratios of the length and duration of successive trends in the Dow Jones Industrial Average cluster around round fractions or Fibonacci ratios. Their study 'Magic numbers in the Dow' covers daily observations on the index over 22,194 trading days, from January 1915 to June 2003, with turning points identified by an explicit procedure borrowed from business cycle analysis, and a test using a block bootstrap. Their conclusion: there is no significant difference between the frequencies with which these ratios occur in cycles in the Dow and the frequencies one would expect to occur at random in such a series; a few significant ratios appear, but no more than would be expected by chance given the large number of tests. The study tests the ratios, not the five-and-three count: it says nothing about the predictive value of a wave 3 or of an expanded flat.

The waves, tested on metals. Marañon and Kumral published in 2018, in the journal Resources Policy, a study exploring the Wave Principle on metal commodity price cycles. According to its abstract, they ran a Monte Carlo simulation over detected Elliott waves in the prices of gold, silver and copper and in a metal price index, and conclude that the Wave Principle would not be a strong approach to analyse commodity markets on a cycle basis, while judging that there is evidence that mass psychology affects those markets, as Elliott posited. That is a precise population, three metals and one index, and the abstract gives no success rate.

One currency, counted after the fact. The article by D'Angelo and Grimaldi cited above concludes that Elliott's model was a valuable tool for predicting the currency market over the period 2009 to 2015. What it actually contains: a wave count laid by the authors on the weekly and daily charts of the euro against the dollar from 2009 to 2017, taken from a charting platform, and one forecast dated 7 May 2014, announcing a fall from 1.40 towards a target of 1.041, when the rate reached 1.0460 on 13 March 2015. There is no counting rule written in advance, no series of forecasts counted, no error measure: a successful count on one pair and one period is an example, not a measurement. And the article proposes two scenarios for what follows, a preferred one and an alternative, which is how a count gets revised.

What no study found measures. We found no published study that fixes a counting rule in advance, applies it to a defined sample, and measures how often the announced wave occurs. The success rates in circulation for Elliott waves therefore have, to our knowledge, no verifiable population, and this lesson repeats none of them. What research establishes about price patterns in general applies here too: Lo, Mamaysky and Wang, in the Journal of Finance in 2000, start from the observation that the presence of a shape on a chart is often in the eyes of the beholder, and have an algorithm recognise ten classic patterns on U.S. stocks from 1962 to 1996. Elliott waves are not among them.

The limits of Elliott wave theory

A subjective count. Batchelor and Ramyar write it when they come to date their peaks and troughs: a technical analyst would do this by eyeballing the chart, marking trends with a ruler or the line drawing tool of some software, and even a simple procedure requires some subjective judgment about what constitutes a reasonable decomposition. Choosing which highs count is already choosing the count. Two honest readers place two different counts on the same chart, and both respect the three rules.

Several valid counts at the same time. Rules exclude, they do not designate. At any given moment, a 'preferred' count and an 'alternative' count coexist, and it is the market that settles it afterwards. The difficulty does not come from a lack of practice, it comes from the very structure of the theory: the same move can be a wave 1 of a higher degree, a wave B of an expanded flat or a wave X of a combination, and the choice is only made after the next wave.

A count revised after the fact. When the market contradicts a count, the theory is not refuted, the count is, and it is replaced by another that respects the rules. One degree more or less, a combination instead of a zigzag, and the whole chart is reread. That is what makes the method so hard to test: a theory that explains everything after the fact predicts nothing beforehand, or at least nothing that can be counted. The two scenarios in the article on the euro against the dollar illustrate it.

What remains useful. The three rules are an honest filter: they say that a reading is wrong, which is more than most chart readings do. And the vocabulary, impulse, correction, expanded flat, describes precisely shapes you will see: the breakout that does not hold, the three-step correction, the triangle before the final move. The article on trading patterns shows how a shape becomes a measurable hypothesis.

In your journal. Tradoshi does not count waves: the count is yours, laid on your chart, with your rule. The journal serves the next step. You write your count before entering, 'short at B of an expanded flat, stop above B', you tag the trade, and the dashboard filters by tag to say what that count has actually given you, personally, on your market. That is the only measurement of a count's predictive value that exists to our knowledge, and it only exists if you make it.

Frequently asked questions