Elliott Wave, Honestly: The 3 Rules, Multiple Counts and Stop Placement

Elliott Wave is the idea that price moves in a repeating rhythm: 5 waves with the main trend, then 3 waves back. If you count correctly, you know which part of the cycle price is in. This article explains the structure and the rules in detail, then covers the part most textbooks play down: the same chart can always be counted more than one way, and that weakness can be turned into a risk management tool.

Where Elliott Wave comes from

Ralph Nelson Elliott was an American accountant. In the 1930s he studied decades of US stock index data and concluded that prices moved in repeating patterns. He published The Wave Principle in 1938 and Nature’s Law in 1946. The idea became popular again when A. J. Frost and Robert Prechter wrote Elliott Wave Principle in 1978, which is still the main reference today. The rules and guidelines in this article follow that book.

The basic shape: 5 waves out, 3 waves back

• Impulse has 5 waves, labelled 1 2 3 4 5. Waves 1, 3 and 5 move with the main trend. Waves 2 and 4 pull back against it.
• Correction has 3 waves, labelled A B C. Waves A and C move against the main trend. Wave B bounces back.

The key point is that the shape repeats at every size (fractal). Waves 1, 3 and 5 each break into 5 smaller waves, and waves 2 and 4 break into 3. Elliott named many levels of wave size (degree), from multi-decade waves down to minutes. In this article we use 1–5 for the larger waves and (i) (ii) (iii) for the sub-waves inside them.

The three hard rules

In a standard impulse there are only three rules that can never be broken. If a count breaks any of them, that count is invalid and you have to count again.

Rule	What it means	If broken
1. Wave 2 never retraces more than 100% of wave 1	In an uptrend, wave 2 cannot go below the start of wave 1	What you called wave 1 is not wave 1
2. Wave 3 is never the shortest	Compare waves 1, 3 and 5. Wave 3 cannot be the shortest (it does not have to be the longest)	Move where you placed wave 3 or 5
3. Wave 4 does not enter wave 1’s price territory	In an uptrend, the low of wave 4 stays above the top of wave 1	That leg is not wave 4, or the move is not an impulse

Rule 3 has one exception: the diagonal (a wedge shape), where the book allows wave 4 to overlap wave 1. It usually appears in the position of wave 1, wave 5, wave A or wave C. This exception makes counting even more flexible, which, as you will see, is both a strength and a problem.

Guidelines, which are not rules

Beyond the three rules, the book lists guidelines: things that “often happen”, not things that “must happen”. Many traders use them as if they were rules.

• Alternation: if wave 2 is a sharp, fast pullback, wave 4 tends to be sideways and slow, and the other way round.
• Wave 3 often extends: it is often about 1.618 times wave 1, or longer.
• Pullback depth: wave 2 often retraces 50% or 61.8% of wave 1. Wave 4 often retraces about 38.2% of wave 3.
• Waves 1 and 5 tend toward equal length when wave 3 is the extended one.
• Channelling: draw a line from the end of wave 2 to the end of wave 4, then a parallel line through the top of wave 3, to see roughly where wave 5 may reach.

These ratios come from Fibonacci, which is where to be careful. In our backtest research on gold (XAUUSD) M1, 06:00–09:00 Thai time (UTC+7), from July 2022 to August 2026, using 1-minute price data and 192,728 bars, price sat near Fibonacci levels about as often as chance would give (a ratio of 0.95–1.00 against the control across 6 swing frames). When price was on a level, the chance of bouncing half an ATR before extending another half ATR was 49.68–50.19%, close to a coin toss. The details are in our article on hindsight bias. That research measured Fibonacci levels, not Elliott Wave directly, and past backtest results do not predict future results. But it is enough to say that a wave ending on a neat ratio is not evidence that the count is right.

The three main corrective patterns

Corrections are much harder to count than impulses, and the book itself admits it. There are three main patterns.

Pattern	Sub-wave structure	What it looks like
Zigzag	5-3-5	A steep pullback. Wave B bounces only a little, and wave C usually goes beyond the end of wave A
Flat	3-3-5	Sideways. Wave B bounces back close to the start of wave A, and wave C usually ends near the end of wave A or a little beyond
Triangle	3-3-3-3-3 (A–E)	A narrowing range. It usually appears in the position of wave 4 or wave B

There are also combinations, such as the double three, which joins two corrections together. The more patterns there are to choose from, the easier it is to find a count that “fits” the chart.

The part rarely discussed: one chart, several counts

The three rules work like a sieve. They remove wrong counts but do not tell you which of the remaining ones is right. Take this example. We built a made-up uptrend with turning points 100 → 110 → 104 → 118 → 113 → 130, and price has now pulled back to 121.

Count A: 100→110 is wave 1, 104 is wave 2, and the whole leg 104→130 is wave 3 (with (i) at 118 and (ii) at 113 inside it). Price is now in wave 4.

• Rule 1: wave 2 retraces (110 − 104) ÷ (110 − 100) = 60%, not more than 100%. Pass.
• Wave 3 is 130 − 104 = 26 long and wave 1 is 10, so wave 3 is clearly not the shortest. Pass.
• Rule 3: wave 4 cannot fall into wave 1’s territory. Invalidation = 110.

Count B: 100→110 is wave 1 and 104 is wave 2, as before. But 104→118 is sub-wave (i) of wave 3, 113 is (ii), and 113→130 is (iii). Price is now in (iv) of wave 3, and wave 3 is not finished yet.

• (ii) retraces (118 − 113) ÷ (118 − 104) = 35.7%. Pass.
• (iii) is 17 long, more than (i) at 14. Pass.
• (iv) cannot fall into the territory of (i). Invalidation = 118.

Both counts pass every rule. The only difference is one question: do the small swings inside the 113→130 leg already contain sub-waves (iii), (iv) and (v)? That depends on the eye of whoever is counting. There is also a count C that passes the rules: 104→118 is wave 3 (14 long), 113 is wave 4 (above the top of wave 1 at 110), and 113→130 is wave 5 (17 long). Wave 3 is not the shortest because wave 1 is shorter. This count says the uptrend has already finished its 5 waves, and the current pullback is wave A of a correction.

The Elliott Wave rules can throw out wrong counts, but they cannot pick the right one. Until price reaches an invalidation level, every count that passes the rules has an equal claim to being right.

Counts revised after the fact, and why testing is hard

As price moves on, counts get revised. If price drops to 116, count B is dead at once, and its supporters will say “it was really count A”. If 110 breaks, count C is left, or the count moves to a diagonal, which allows waves to overlap. Each revision looks neat and correct on the historical chart, because it was drawn after the answer was known. It is similar to indicators that repaint: the last wave is never confirmed until it is already over.

This is why Elliott Wave is very hard to backtest in a systematic way.

• There is no single count. To let a computer count, you have to write rules for how big a swing must be to count as a wave, and which count to choose when several pass. Every set of rules is a “different version of Elliott Wave”.
• There are many escape routes. Diagonals, combinations and shifting the wave degree mean almost any chart can be given a count that fits. A theory that can explain everything predicts very little.
• Try many, keep the best. If you test a hundred sets of counting rules and keep the one with the best results, you run into data snooping. A good result may be pure luck.

So there is no clear evidence that the idea predicts direction. That does not make it useless.

Using Elliott Wave as a risk framework

The part that does work is this: every count has a clear invalidation level that comes straight from the rules. If price reaches it, the count is definitely wrong, with no interpretation needed. That makes it a stop loss position with a reason behind it. Professional wave analysts don’t hold a single count either. They keep a preferred count and an alternate count, each with its own invalidation level.

Back to the example. Suppose a trader who thinks the uptrend is not over buys at 121, is willing to lose at most $100 (1% of a $10,000 account), and places the stop 0.5 below the invalidation level. Assume a 1-point move is worth $1 per unit.

	Count A	Count B
Invalidation level	110	118
Stop loss	109.5	117.5
Stop distance	121 − 109.5 = 11.5	121 − 117.5 = 3.5
Position size = 100 ÷ distance	8.70 units	28.57 units

The money at risk is the same, $100, but the position sizes differ by 11.5 ÷ 3.5 = 3.29 times. The general method is in position size from your stop loss, and you can work it out with the lot size calculator.

Simple rules of thumb:

1. Write down the invalidation level before you enter. If your count has no clear invalidation level, the count isn’t finished. Don’t enter yet.
2. Set the stop from the count you are using, not from hope. Count B gives a bigger position, but if 118 breaks you must get out, even though count A is still alive. Don’t move the stop to 110 as price gets close.
3. Look at the count that makes you wrong. Count C says the uptrend is over. If it passes the rules, you have to accept that the chance you are wrong is real.
4. Use RR from the invalidation level. Compare the stop distance with a reasonable target using the RR calculator. If it only pays off with a very distant target, skip the trade.

Use Elliott Wave to ask “at what price will I know I’m wrong?”, not “where will price go?”

Common mistakes

• Treating guidelines as rules, for example throwing out a count because wave 2 didn’t reach 61.8%. The book never says it must.
• Clinging to one count and revising it every time price disagrees, while never leaving the trade.
• Moving the stop to a new count. Once price breaks the invalidation level, the original count is wrong. Hunting for a new count so you can hold on is moving your stop loss in disguise.
• Counting on too small a timeframe. Small charts have so many little swings that almost any count fits. Start on the bigger chart.
• Trusting neat historical examples. The finished counts in articles and videos were mostly drawn after the outcome was known. Write down your own counts and invalidation levels in advance, then see how often they hold.
• Forgetting position size. A closer stop means a bigger position for the same money at risk. Recalculate every time you change your count.

If you are interested in other ways of reading price structure that have been put to the test, read on with Smart Money Concepts and Wyckoff Power of Three.