The Lab

What Elliott traders look at

Elliott wave theory says markets move in a repeating shape.

• The impulse: five waves with the trend. Waves 1, 3 and 5 go with it. Waves 2 and 4 are pullbacks against it.
• The correction: three waves against the trend, called A, B and C. Or a triangle of five shrinking waves, A-B-C-D-E, before the trend carries on.
• Waves within waves. Each wave is made of smaller waves of the same shape, and is itself part of a bigger one. A wave 3 on the daily chart is five smaller waves on the hourly chart.

The three hard rules. A count that breaks one is wrong:

1. Wave 2 never falls below the start of wave 1.
2. Wave 3 is never the shortest of waves 1, 3 and 5.
3. Wave 4 never falls into wave 1's range.

The zones where waves are said to turn. These are guidelines, not rules. They use the Fibonacci ratios:

• wave 2 ends 50% to 61.8% of the way back down wave 1;
• wave 4 ends about 38.2% back down wave 3;
• wave 3 runs about 1.618 times wave 1;
• wave 5 is about as long as wave 1;
• a correction tends to end near the previous wave 4;
• in a correction, B goes 50% to 61.8% back up A, and C is about as long as A.

The trades they take: long when wave 2 ends (to ride wave 3), long when wave 4 ends (to ride wave 5), short when wave 5 ends (to ride the correction). Some also short waves 2 and 4 themselves.

How we tested

• Prices: hourly Bitcoin, Ether and Solana futures, January 2021 to September 2026. The first half (to October 2023) was used to pick settings, and the second half to judge them.
• Counting the waves as they happen. Elliott waves are usually counted on a finished chart, where every turn is already known. That is the classic trap. Here a turn counts only once the price has moved far enough away from it to confirm it, as a trader would see it at the time.
  • "Far enough" depends on how much the coin has been moving lately.
  • Three sizes, the waves within waves: small (one day's typical move), medium (2.5 days') and large (6 days').
• The labels. At each confirmed turn, the turns before it were checked against the three hard rules, and the turn got its wave name.
• One count. Elliott often allows more than one count at once. We kept the one using the most waves (the fewest still open), then the one with the biggest wave 3.
• The zones. For each zone, we took the share of turns that ended inside it. We compared that with 4,000 random zones of the same width, placed where turns of that wave actually happen. This was done both as counted at the time and with hindsight (waves that a later count confirmed).
• The trades: seven of them, with the count's own invalidation level as the stop (for example, wave 1's start for a long at the end of wave 2) and the zone as the target.
  • Fees were counted; funding wasn't.
  • Each was compared with random entries in the same month, with the same stop and target distances.
• The bar to pass, set before the run. A zone or trade passed only if it beat random by a wide margin, allowing for how many tests were run (a Bonferroni correction). For the zones it also had to hold in both halves.

The maths, step by step

1. A wave-2 zone. Wave 1 runs from 100 to 110, so it is 10 long. Wave 2 ends at 104, which is 6 back from 110:

6 ÷ 10 = 60%. That is inside the 50%–61.8% zone.

2. Is the zone special? Say 20% of all wave 2s end in the 50%–61.8% zone. Now slide a zone of the same width (11.8 points) to random places where wave 2s end, for example 30%–41.8%, or 65%–76.8%. If many of those zones also hold 20% or more, the Elliott zone isn't special. It's just where wave 2s often are.

3. Why wave 4 looks so good, and isn't. Wave 4 may not fall into wave 1's range. So it can only go back so far: often less than half of wave 3. That squeezes nearly every wave 4 into a narrow range. A 38.2% zone then catches about half of them. Any zone of the same width inside that range does about as well.

4. A trade. Long at the end of wave 2 at 104, with the stop at wave 1's start (100) and the target at 104 + 1.618 × 10 = 120.2.

• Risk: 4. Reward: 16.2.
• To break even, before fees, it must win about one time in five (4 ÷ 20.2 = 20%).

What we found

1. The count is often open. At 16% to 28% of turns, more than one Elliott count fit the rules. Even with the rules, the count is a choice about one time in five.

2. The zones are no better than random ones. In none of the tests did an Elliott zone hold more of the turns than other zones of the same width. That was true counted at the time and with hindsight, on all three coins and all three sizes.

• 42 tests had enough turns in both halves; the rest had too few, mostly at the large size.
• Wave 2s ended in the 50%–61.8% zone 12% to 32% of the time. Other zones of the same width did about as well.

3. The fib question. Our earlier Fibonacci study found no edge in the fib levels on their own. Does the wave context help?

• With hindsight, wave 4s ended in the 38.2% zone 40% to 58% of the time. All swings, without wave labels, did so only 5% to 7% of the time.
• That gap is the rules at work, not an edge. As in step 3 of the maths, the rules squeeze wave 4 into a narrow range. Inside that range, the 38.2% zone is ordinary.
• The wave context adds nothing beyond the fib levels.

4. The trades don't beat random. None of the seven trades passed, on any size.

• Longs, waves 3 and 5: at or below random entries on every size.
• Shorting waves 2 and 4, medium size: the closest to passing.
  • Shorting wave 2: +0.50% a trade, against −0.85% for random. It made money in both halves, but wasn't strong enough to pass.
  • Shorting wave 4: +0.19% against −1.11%, but it lost in the first half.
• Shorting wave A (after wave 5), and the triangle thrust: at or below random.

5. Waves within waves don't line up. A small count agreed with the medium count, and the medium with the large, 43% to 46% of the time. With the larger count shifted to random times, they agreed 50% of the time. The sizes don't line up more than chance would give.

Credits

• Ralph Nelson Elliott, The Wave Principle, 1938, and Nature's Law, 1946: the waves, and their link to Fibonacci ratios.
• A. J. Frost and Robert R. Prechter Jr., Elliott Wave Principle, New Classics Library, 1978: the rules and guidelines as tested here.
• Leonardo of Pisa (Fibonacci), Liber Abaci, 1202: the number sequence behind the ratios.
• Carlo Emilio Bonferroni, 1936, and Olive Jean Dunn, 1961: raising the bar for many tests.
• The live wave counter, the random zones and the trade test are our own design. The fib comparison builds on our study "Fibonacci confluence: does it time trades better than chance?".
• Data: Binance's public hourly futures prices.
• PerpGremlin and Claude, Elliott waves: do the zones and the wave trades hold up?, TerminalGremlin, 2026: the tests, the code and the words.