What Is Elliott Wave Theory? The 3 Rules, Tested on 33 Years of SPY

Michael Whitman · Trading since 2017, full-time since 2020 · YouTube creator, 137K subscribers · Published 2026-09-05 · Updated 2026-11-02 · Reviewed 2026-11-01

Covered on this page: TradingView.

Elliott wave theory says a trending market moves in a repeating eight-part rhythm: five waves in the direction of the trend, then three waves against it. Ralph Nelson Elliott described it in the 1930s, and A.J. Frost and Robert Prechter brought it to a wide audience in 1978. Most explanations stop at the diagram.

This one also checks the theory’s three rules against every swing on SPY’s daily chart since 1993, and walks through a real count from 2024 and 2025 with the prices on it.

How it works

The five waves with the trend are called the impulse. Waves 1, 3 and 5 move with the trend. Waves 2 and 4 are pullbacks inside it. In a rising market, wave 1 is the first push up, wave 2 gives some of it back, wave 3 carries price past the top of wave 1, wave 4 pauses, and wave 5 makes the last new high.

The three waves against the trend are the correction, labeled A, B and C. In a rising market, A falls, B bounces part of the way back, and C falls again. Together, five plus three is one full cycle. Corrections come in three main shapes: the sharp zigzag, the sideways flat, and the triangle, which has five overlapping legs labeled A to E instead of three.

The pattern is said to repeat at every scale. Each impulse wave splits into five smaller waves, and each corrective wave into three, so a wave 3 on a weekly chart is a full five-wave move on a daily one. Elliott named these scales “degrees”, from cycles lasting decades or longer down to moves lasting minutes.

The idea behind the shape is crowd behavior: optimism builds in waves, gets ahead of itself, and unwinds in a pattern of its own. That is the theory’s explanation, not something a chart can prove.

The three rules, and the guidelines that are not rules

Everything that can make a count wrong comes down to three statements about an impulse:

The one standard exception is the diagonal, a rarer wedge-shaped five in which wave 4 is allowed to overlap wave 1. The table, and the test further down, cover the ordinary impulse.

The guidelines are the long list that starts with “often”. Wave 2 often retraces 50% to 61.8% of wave 1. Wave 4 is often shallower than wave 2. Waves 2 and 4 often alternate between a sharp drop and a sideways drift. Wave 5 often travels about as far as wave 1. Most of these come from Fibonacci ratios, which is why the two subjects usually appear together.

The difference matters more than it looks. A count that ignores every guideline is still a legal count, because none of them can disqualify anything. Only the three rules can. When a chart shows a wave count covered in retracement percentages, the percentages are describing the count rather than deciding whether it stands.

A worked example: SPY, April 2024 to April 2025

The swings below come from a mechanical filter, not from eyeballing: a new swing is marked only after SPY moves 5% the other way from its last extreme, using daily highs and lows. That rule is the same one behind a 5% zigzag indicator, and it picks the turns before any labeling starts.

The count. Label the swings from the 19 April 2024 low as a five-wave advance:

Check the rules in order. Rule one: wave 2 bottomed at $510.27, above wave 1’s start of $493.86. Rule two: wave 3 covered $98.80, more than wave 1’s $71.30 and wave 5’s $37.88, so it was the longest. Rule three: wave 4 bottomed at $575.35, $10.19 above wave 1’s top of $565.16. The count passes all three. It also ignores two guidelines: wave 2 went deeper than 61.8%, and wave 5 was about half the length of wave 1. Neither changes its legality.

The rules also give the count a price at which it dies. While wave 5 was forming, a trade below $565.16 would have put wave 4 inside wave 1’s range and broken rule three. That is the most useful number the theory produces, because it is fixed in advance and does not depend on anyone’s opinion.

What happened next. SPY’s low reached $582.44 on 28 February 2025, more than 5% below the $613.23 top, which is the first day the filter could call wave 5 finished. The close that day was $594.18. The next swing low came on 7 April 2025 at $481.80, 21.43% below the top. On this count, that drop was the correction after the five.

Now the catch. The same bars carry a second legal count. Start from SPY’s 13 March 2023 low of $380.65 and the advance to 16 July 2024 also passes all three rules: wave 1 to $459.44, wave 2 to $409.21, wave 3 to $524.61, wave 4 to $493.86, wave 5 to $565.16.

On that reading, July 2024 was the end of a five-wave advance and a correction was due. On the first reading, July 2024 was only wave 1, with the strongest leg still ahead. SPY then climbed to $613.23. Both counts obey every rule, which is the point the first-person note above makes.

Why one chart can hold several counts

Three rules on eight labels leave a lot of room. The rules throw out some labelings, but they rarely leave just one. The example shows the most common version of the problem: a high that ends a completed five at one degree is the first wave of the next five at another.

The fractal design adds more room. If a move labeled wave 3 turns out to be wave 1 of something larger, the count is not broken, it has moved up a degree. A theory that can absorb that outcome has given back some of the testability its rules bought.

And counts get relabeled as bars arrive. A drop that breaks rule three does not end the analysis. It forces a new count on everything before it. So a live count is a current best guess, not a fixed object, and acting on it means acting on something that can change backward.

The practical answer is to write the alternative down. For the July 2024 high that meant two lines: “five complete, correction due” and “wave 1 done, wave 2 under way, invalid below $493.86”. Whichever one price breaks first decides which reading survives, and nobody has to guess which was right beforehand.

The original data

The test. SPY daily bars from 29 January 1993 to 30 October 2026, 8,497 sessions, raw prices. The 5% swing filter from the worked example found 402 swings. Every stretch of six alternating swings that starts at a low (low, high, low, high, low, high) is one candidate five-wave advance, 198 in all. Each was checked against the three rules exactly as written in the table above. The windows overlap, so they are not 198 independent events.

How often each rule held:

Rule three does almost all the filtering. On a 5% filter, a pullback that stays above the top of the previous push is uncommon, so about nine in ten runs of five swings on SPY are not a valid impulse. Of the 18 that passed, 13 had wave 5 above wave 3 and 5 ended with a lower wave 5 (a “truncated” fifth, which the rules allow).

Did a valid count tell you more about what came next? The theory says a correction follows a completed five. After the 18 rule-passing advances, the next swing low was a median 7.57% below the wave 5 top. After the other 180 runs it was 7.80%. Those are not different in any meaningful way (Mann-Whitney test, p = 0.27).

Sixty sessions after the day each fifth wave was confirmed, SPY was higher after 14 of the 18 (77.78%), against 5,919 of 8,437 sessions of all kinds (70.16%), a gap well within chance (p = 0.61). Eighteen is a small sample, and the honest summary is that a legal count did not separate what followed from what followed any other swing high.

The 50% to 61.8% guideline for wave 2 matched 3 of the 18 valid counts, and 15 of all 198 runs. The median wave 2 in a valid count retraced 65.45% of wave 1.

The filter size did not change the picture. With a 3% filter, 33 of 430 runs (7.67%) passed all three rules; with 8%, 8 of 71 (11.27%). In both, the drop after a valid five was within chance of the drop after any other run. Every run, its prices and its rule results are in the SPY Elliott rule file.

What people publish about it. Across our study of 24,971 trading videos, 90 have Elliott wave in the title, with a median of 5,502 views. Of the 78 with description text, 1 mentions invalidation or what a wrong count looks like, on the one method in this glossary that comes with written rules for being wrong.

When it fails

Most stretches have no valid count

On SPY’s daily swings, nine runs in ten broke at least one rule. A chart without a clean impulse is the normal case, and forcing labels onto it is how a count turns into a story.

The degree is always negotiable

A failed count can be moved up or down a degree and declared intact. That escape route is open every time, which is why the rule-three price matters more than the label.

The count only looks obvious afterward

Every finished five in the table above was confirmed after its top, not at it. In the worked example, the filter could call wave 5 complete on 28 February 2025, by which time SPY was already 3.11% below the high at the close.

Two counts, opposite plans

In July 2024 one legal count said “sell the end of the five” and another said “buy wave 2”. The rules had no way to pick between them, and only later prices did.

Questions people ask

What is Elliott wave theory in simple terms?

It is the idea that prices trend in five steps forward and three steps back, and that the same pattern shows up on every timeframe. Analysts label the swings 1 to 5 and A to C to decide where in the cycle the market sits.

What are the 3 rules of Elliott wave?

Wave 2 cannot fall below the start of wave 1, wave 3 cannot be the shortest of waves 1, 3 and 5, and wave 4 cannot overlap the price range of wave 1. Break any one and the count is wrong.

Is Elliott wave theory reliable?

In this test, a count that passed all three rules was not followed by a bigger drop or a different next 60 sessions than any other swing on SPY, and the results were within chance. Its clearest use is setting a price where a reading is proven wrong, not forecasting.

What comes after wave 5?

The theory says a three-wave correction, A, B and C. On SPY since 1993, the drop after a rule-passing fifth wave was about the same size as the drop after any other swing high.

Related

Market structure reads the same highs and lows without labels, so there is no count to disagree about. Swing highs and lows decide which turns are big enough to be waves at all. Fibonacci supplies the ratios in the guidelines, and has the same choose-your-swing problem. For the step-by-step labeling process, see how to count an Elliott wave.

This page is educational, not financial advice. Test every idea on your own charts before risking money. Some links on this page earn a commission if you buy through them. It costs you nothing and it does not decide what appears here or in what order — how these pages are made is set out in our methodology.