Fibonacci Retracement: what it is, and whether it works

804 of 3,497 out-of-sample tests beat simply buying and holding the same asset — 23.0%. On the other 2,693 it did not. That is above the 20.1% rate across all 382 indicators we test (one pooled rate over all 660,005 tests we have run, not a mean of the per-indicator rates).

One test is one variant on one asset on one timeframe. We tested 2 variants of Fibonacci Retracement, so an asset that carries several of them contributes several tests — this is the family's record across everything we ran, not a count of assets. There are only two, and they differ: 22.1% and 24.0%. 23.0% is the pooled rate over both, not a midpoint and not a property either one has on its own — the table below gives each separately. "Beat" means a higher return than holding that same asset over that same window. Measured out-of-sample, on data the setup was not chosen on.

A swing high and a swing low are chosen, and the vertical distance between them is divided at fixed proportions. 61.8% is the inverse of the golden ratio, 38.2% is that figure squared, and 23.6% is its cube; 78.6% is the square root of 61.8%. 50% is not a Fibonacci proportion at all but is conventionally included. The levels are entirely a function of which two points are selected — the arithmetic adds no information beyond that choice.

How it's read. The levels are read as candidate support in an uptrend pullback and candidate resistance in a downtrend rally, with 38.2% to 61.8% treated as the zone where a healthy pullback tends to end.

Where it struggles by design. The tool has no objective input: two analysts picking different swing points produce different levels on the same chart, and there is no rule in the method for choosing them. That makes it unusually easy to fit after the fact, and unusually hard to test honestly — which is worth keeping in mind when reading any published result about it, including ours.

Origin: Fibonacci proportions entered technical analysis through Ralph Nelson Elliott, who connected them to his wave work in the early 1940s and set it out in Nature's Law (1946) — by his own account he had never heard of the Fibonacci series when he first described the waves in the 1930s.