Fibonacci retracements are the most popular drawing tool in retail trading and the one with the thinnest theoretical justification. That combination is worth understanding clearly, because the honest case for using them is not the one usually given.
Where the numbers come from
The Fibonacci sequence — 1, 1, 2, 3, 5, 8, 13, 21 — has each term as the sum of the previous two. As it extends, the ratio between consecutive terms converges on 0.618, and a few related ratios fall out of the same arithmetic:
- 61.8% — the ratio of a term to the next one.
- 38.2% — the ratio of a term to the term two places on.
- 23.6% — three places on.
- 78.6% — the square root of 61.8%.
These ratios genuinely recur in natural growth patterns. The leap from that to "a currency pair will retrace 61.8% of its move" is not a derivation, and nobody has supplied the missing step. Better to be straight about that than repeat the sunflower anecdote.
Note too the level traders watch most closely: 50%. It is not a Fibonacci ratio at all. It is on the tool because a halfway retracement is an intuitive idea that long predates any of this — which rather undercuts the mystical framing of the rest.
The honest reason it works at all
Fibonacci levels have some predictive value for the same reason round numbers do: enough participants watch them that orders accumulate there. Retail platforms ship the tool by default, institutions know retail is watching, and algorithms are programmed around the levels.
That is a real effect. But it is a statement about crowd behaviour, not mathematics, and it has a consequence: the levels work best where many people are looking — major pairs, daily and 4-hour charts, recent and obvious swings. On an illiquid cross at 3-minute resolution nobody is watching, and the ratios have nothing left to stand on.
Drawing it without fooling yourself
The tool is anchored between a swing low and a swing high. Everything depends on that choice, and the choice is where the self-deception enters.
- Use the obvious swing. If you have to hunt for the anchor points that make the levels line up with where price already stopped, you are not analysing — you are decorating.
- Pick one convention for extremes — wicks or bodies — and use it every time. Alternating between them roughly doubles the number of "levels" available, which is how a chart ends up with a level near everything.
- Draw it before price gets there. A retracement level identified in advance is a hypothesis. One identified afterwards is a coincidence with good PR.
- Match the swing to your timeframe. Levels from a three-year move are irrelevant to a two-day trade.
The levels in practice
The commonly observed pattern: shallow retracements to 23.6% or 38.2% suggest a strong trend that barely paused, while deep ones to 61.8% or 78.6% suggest a trend under real pressure. A move retracing beyond 78.6% is better described as a reversal than a pullback, whatever the tool is labelled.
The 61.8% level attracts the most attention and therefore the most stop-hunting. Price wicking just beyond it before turning is common enough that a stop placed immediately below a Fibonacci level is close to advertising your exit.
Confluence, and its trap
The defensible way to use the tool is as one input among several: a 61.8% retracement that lands on a prior support level and a rising 50-period moving average is a more interesting area than any of the three alone.
The trap is that confluence can be manufactured. With enough indicators, swings and ratios, some combination aligns near any price you like. It is only evidence if the components were chosen before you went looking.
Extensions
Extensions project beyond the original move — 127.2%, 161.8%, 261.8% — and are used for targets. The same reasoning applies with less force: fewer traders watch them, so the crowd effect that gives the levels their substance is weaker.
What to take from this
- The ratios come from the Fibonacci sequence; the leap to price behaviour is not derived.
- 50% is not a Fibonacci number — it is on the tool because halfway is intuitive.
- The levels work mainly because many participants watch them, so they fade on obscure charts.
- Anchor to the obvious swing, keep one wick/body convention, and draw it in advance.
- Confluence counts only when the components were chosen before the search.
That completes the indicator toolkit. Every tool in it is computed from price, which means none of them can tell you anything price has not already said — their value is in making one aspect of it legible at a glance. The modules ahead turn to the structure those tools are measuring, and then to the macro forces that set the direction in the first place.