Fibonacci retracements & extensions

Definition

Fibonacci levels mark fractional retracements (23.6%, 38.2%, 50%, 61.8%, 78.6%) of a chosen price swing, and extensions beyond it, on the premise that pullbacks and targets cluster at ratios derived from the Fibonacci sequence. The platform documents the construction because it is ubiquitous practitioner vocabulary — and states plainly: there is no peer-reviewed evidence that Fibonacci ratios mark price levels better than arbitrary fractions. This entry’s primary content is that verdict.

How it works / structure

• Construction: choose a swing low and high (the anchoring choice — unstandardized and decisive); divide the range at the ratio set; extensions project the ratios beyond the swing (127.2%, 161.8%).
• Parameters (engine-executable, for testing): swing definition (mechanical extremum rule — mandatory for falsifiability), ratio set, tolerance band around each level, and the claim form (touch-and-hold frequency vs random levels).
• The testable question: do reversals cluster at Fibonacci fractions MORE than at uniformly spaced fractions with the same tolerance? Published attempts find no reliable excess — and 38.2%/50%/61.8% with realistic tolerance bands cover most of a pullback’s plausible territory, making “the level held” nearly unfalsifiable as commonly practiced.

When it applies

As shared vocabulary when analyzing discourse and crowd attention (many participants watch the levels — a self- fulfilling-anchor hypothesis that is itself testable, and distinct from any mathematical property of the ratios); as pre-registered levels in a mechanically-defined replay test. The platform generates no signals from Fibonacci levels.

Risk profile & failure modes

• Unfalsifiability as practiced: free choice of swing anchors plus five levels plus tolerance means SOME level always “worked” in hindsight — the construction retrofits any path.
• Anchoring bias with mathematical costume: the ratios’ Fibonacci provenance lends false authority to what are arbitrary fractions of an arbitrarily chosen range (bias-anchoring).
• Narrative displacement: levels substitute for theses — “bounced at the 61.8%” explains nothing and predicts nothing without a mechanism.

Evidence & limits

The technical-analysis survey literature (Park-Irwin 2007) contains no validated Fibonacci-level edge; targeted academic examinations of ratio-level clustering find none distinguishable from chance. The one defensible hypothesis — crowd-watched levels briefly concentrate order flow — is a self-reference effect shared with round numbers and pivot points, not a property of the golden ratio. Classification: folklore, with the vocabulary documented for interoperability.

Falsifiable-thesis examples

Illustrations only, not signals:

• “Mechanically-anchored 61.8% retracement touches on universe U hold (no close beyond) more often than same-tolerance uniform fractions this replay decade” — falsified by the paired frequencies.
• “X’s current pullback will hold above the 61.8% level of its last mechanical swing” — falsified by a close below.

Cross-references

• Fellow level constructions: indicator-pivot-points (same self-reference hypothesis, same evidence status)
• The bias it dresses up: bias-anchoring
• Discipline: lens-technical (falsifiability requirements)
• Where the vocabulary appears: strategy-swing-trading discourse