How to Count an Elliott Wave
To count an Elliott wave, label the five-wave move in the direction of the trend and the three-wave correction against it, then check the count against the three hard rules. If any rule is broken the count is wrong and has to be redrawn rather than adjusted.
Elliott wave describes trends as a five-wave move followed by a three-wave correction, repeating at every scale. Three rules are strict. Everything else is a guideline, and the difference between those two categories is the whole practical content of the method.
Before you start
The three hard rules written down, because they are the only part that can invalidate a count. Everything else can be argued either way.
One timeframe, chosen in advance, since the pattern is fractal and every scale has its own count. A wave 3 on the hourly sits inside a wave 1 on the daily.
An acceptance that your count will be revised, and a rule for what a revision costs. Unlimited revisions make the framework unable to be wrong.
The steps
1. Write the three rules at the top of the page
Wave 2 never retraces beyond the start of wave 1. Wave 3 is never the shortest of 1, 3 and 5. Wave 4 never enters wave 1’s price territory.
2. Fix the timeframe and the starting point
A count without a stated timeframe and origin is not a claim about anything. Both go in writing before a single label is placed.
3. Label the impulse: five waves with the trend
Three advances separated by two retracements, in the direction of the larger trend. Label them in order and do not skip ahead to the wave you would prefer to be in.
4. Test the count against all three rules
Any one broken means the count is wrong. Not adjusted — wrong. Redrawing from the start is the correct response and it is the step people skip.
5. Label the correction: three waves against
Corrections take many shapes and are considerably harder to label than impulses. Ambiguity here is normal and is not a sign you are doing it wrong.
6. Write the count down with a date
The count, the timeframe, the date, and the price that would invalidate it. Four things. That last one is what turns the exercise into a position with a stop.
7. Trade the invalidation level, not the forecast
Each rule gives a specific price that would prove the count wrong. That level is a stop, and the distance to it sets the size — which is the most useful thing the framework produces.
How to tell it worked
All 3 rules were checked against the count, not assumed.
The timeframe and origin were fixed before any label was placed.
The count was written down with an invalidation price, so it can be wrong.
And a broken rule produced a redraw 100 percent of the time, never a relabel.
What separates the rules from the guidelines
The three rules produce specific prices. They are the only part of the framework that can be falsified by a chart, and they are what makes a count tradeable at all.
The guidelines produce expectations. Wave 3 is often the longest, wave 2 and wave 4 usually differ in shape, retracements often land near certain ratios. All are frequently true and none can invalidate anything.
The revision problem
Counts are routinely revised as new bars arrive, and some revision is legitimate — the framework is explicitly about nested structure, so a higher-degree label can genuinely change.
But an unlimited revision budget removes every possible failure. If any adverse move can be reinterpreted as a different wave at a different degree, the count has stopped making a claim.
A revision budget is the fix, and it is a number you choose. Two revisions per count, say, after which you are flat and start again. On this site’s shared series direction runs average 2.01 bars and the longest ran 11 — the market supplies plenty of movement to reinterpret, and the limit has to come from you.
The one output worth having
An invalidation price, arrived at by a stated method. Whatever you think about the forecasting claims, the three rules produce a specific level at which the count is definitively wrong — and that is a stop with a reason behind it.
Which means the framework is usable even if you doubt the rest of it. Count the structure, take the rule-derived level as your invalidation, size the position from that distance, and ignore every projection about where wave 5 ends.
That version has an honest failure mode. The level breaks, you are out, and the count gets redrawn. It costs one defined loss rather than a sequence of relabels.
And it is testable in a way the forecasts are not. After thirty counts you can say what proportion of your invalidation levels held, which is a number — unlike the question of whether the market really moves in fives.
The original data
Of the 24,971 unique videos in research/search-study-corpus.jsonl, 90 mention this framework in the
title, at a median of 5,502 views across 38 channels — and only 22% of those titles are
instruction-shaped, the lowest proportion of any subject measured on this site. Order blocks appear in
391 at 2,786. The counts come from site/corpus_count.py.
22% instruction-shaped against 61% for the neighbouring frameworks. People arrive at this one wanting to understand it rather than to apply it, which is an unusual and honest signal about how tractable the method actually is.
The answer to the question on that chart is that a hard rule was broken. The count is wrong and the correct response is to redraw from the origin — relabelling the whole sequence one degree down to preserve it is exactly the move that makes the framework unfalsifiable.
When it fails
The failure is the count that survives everything, and it takes months to notice. Price moves against the forecast, so the sequence is relabelled one degree lower. It moves further, so an extended wave is introduced. Each revision is permitted by the literature and each preserves the original directional view. A year later the count has never been wrong, has never produced a losing trade on paper, and has never once said anything that could have been checked in advance.
The second failure is no stated timeframe. The pattern is fractal.
A third is relabelling after a rule breaks. That is the falsifiability leaving.
A fourth is trading the forecast rather than the invalidation. Only one of those is a price.
A fifth is counting corrections with impulse confidence. They are much harder.
And a sixth is no revision limit. Without one the count outlives every piece of evidence.
Related
Elliott wave covers the structure and the full rule set. Retracement is the measurement the guidelines lean on. And trend analysis is the simpler framework this elaborates.
The three rules are what make this worth anything, because they are the only part that can tell me I am wrong. Everything else — the ratios, the extensions, the alternation guidelines — bends. I write the count down with a date on it, and if a rule breaks I start again rather than relabelling until it fits.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.