Research
Do crypto prices turn at Fibonacci levels?
Traders draw Fibonacci retracements on almost every chart: after a move, price is said to pull back to 38.2%, 50% or 61.8% of it and turn there. We measured 11,088 completed retracements on 94 coins and checked where each one actually ended.
No better than chance
None of the five levels attracted more endings than the prices right beside it, at any of the three swing sizes examined. The largest excess anywhere was 40% of nearby endings against an expected 33%, which chance produces easily across 15 comparisons.
Level by level
| Level | Ended within 2 points | Ended just beside | Share at the level | Expected | By luck |
|---|---|---|---|---|---|
| 23.6% | 127 | 293 | 30% | 33% | 95% |
| 38.2% | 401 | 843 | 32% | 33% | 95% |
| 50% | 590 | 1,169 | 34% | 33% | 93% |
| 61.8% | 744 | 1,425 | 34% | 33% | 48% |
| 78.6% | 678 | 1,406 | 33% | 33% | 95% |
For each level, “ended within 2 points” counts retracements that stopped between 2 points below and 2 points above it. “Just beside” counts those that stopped in the two strips of the same width on either side. If the level meant nothing, one ending in three would land in the middle strip. “By luck” is how often chance alone gives a share at least that high, corrected for all 15 comparisons on this page.
If you waited at the level
The trader's version of the question: of the retracements that got as far as a level, how many stopped there? Deeper levels always hold more often, because fewer moves are left to pass through them. So a Fibonacci level has to be judged against the plain levels around it, which have no special meaning.
| Level | Kind | Reached it | Stopped within 2 points | Share |
|---|---|---|---|---|
| 23.6% | Fibonacci | 10,961 | 127 | 1.2% |
| 30% | plain | 10,712 | 238 | 2.2% |
| 38.2% | Fibonacci | 10,111 | 401 | 4.0% |
| 44% | plain | 9,498 | 527 | 5.5% |
| 50% | Fibonacci | 8,716 | 590 | 6.8% |
| 56% | plain | 7,827 | 667 | 8.5% |
| 61.8% | Fibonacci | 6,845 | 744 | 10.9% |
| 70% | plain | 5,304 | 694 | 13.1% |
| 78.6% | Fibonacci | 3,751 | 678 | 18.1% |
| 86% | plain | 2,509 | 621 | 24.8% |
Pullbacks and bounces separately
| Level | Pullbacks: share at the level | Bounces: share at the level |
|---|---|---|
| 23.6% | 19% | 31% |
| 38.2% | 32% | 32% |
| 50% | 33% | 34% |
| 61.8% | 33% | 36% |
| 78.6% | 30% | 36% |
A pullback is the fall that follows a rise; a bounce is the rise that follows a fall. Expected share with no effect: 33%.
Smaller and larger swings
| Swing size | Retracements | 23.6% | 38.2% | 50% | 61.8% | 78.6% |
|---|---|---|---|---|---|---|
| 10% or more | 26,978 | 32% | 34% | 35% | 35% | 32% |
| 20% or more | 11,088 | 30% | 32% | 34% | 34% | 33% |
| 30% or more | 5,944 | 40% | 33% | 33% | 35% | 32% |
Share of nearby endings that fell at each level, for three definitions of a swing. Expected with no effect: 33%.
How it was measured
A swing high is confirmed once price has fallen 20% from it, and a swing low once price has risen by the same factor, using each day's high and low. Three swing points in a row give a move and its retracement. The depth is the retracement divided by the move: a rise from $100 to $200 followed by a fall to $150 is a 50% retracement. Retracements between 10% and 100% are kept. The same rule runs on every coin with no manual chart reading.
Limits: daily candles only, so levels touched and left within a day are judged by that day's extreme. A retracement only counts once it has ended, which is how a chart reader sees it in hindsight too. This says nothing about whether a level is useful as a place to put an order, only that price does not turn there more often than nearby.
Open data for this page · How the rest of the site tests signals · Signals that did hold up
Every night the signals that fired are sealed with their history. Members see the list straight away.