4 / 4 · 6 min
Fibonacci: what here is measurable and what is decoration
There is no law of nature obliging price to respect 0.618. There is something else — and it is enough to make the levels work, but not in the way they are advertised.
The ratios 0.382, 0.5, 0.618 come from a number sequence that has nothing to do with markets. There is no mechanism forcing price to stop exactly there, and there could not be one. But there is a simpler mechanism: millions of people and trading programs draw those levels, and other people's orders pile up on them. Exactly the same nature as head and shoulders: what works is not the prediction but the fact that people watch it.
What follows from that in practice
A Fibonacci level is an ordinary level and must be treated as one from the lesson on levels: it is a ZONE, not a line, and on its own it decides nothing. It becomes useful when it coincides with something that has a mechanism of its own: last week's high, a round number, a density in the order book. A lone 0.618 line in the middle of an empty chart is decoration.
Marking a pullback out, in numbers
The impulse ran from 1,480 to 2,260 — 780 points. Retracements are measured from the top: 0.382 gives 1,962, 0.5 gives 1,870, 0.618 gives 1,778. Three numbers, all three legitimate — which by itself ought to be a warning. If you have three levels across a two-hundred-point range, price will always end up near one of them and an explanation will be found after the fact for any outcome. It only makes sense to work with ONE chosen in advance, and only if there is something else beside it.
Anchoring the grid at arbitrary points
Everything in this construction is decided by two anchor points, and you pick them. Move the start by one candle and every level shifts. On finished history a person unwittingly places the anchors so the lines coincide with reversals that already happened: that is not markup but fitting to a known answer. The rule: anchors are taken by an objective sign — the start and end of a wave in your own structure markup — and are not moved once placed.
Drawing several grids at once
Two or three grids from different waves give a dozen lines on the screen, and then the same thing happens as with twenty levels: price is always near one of them. Several lines meeting at one point is presented as a stronger signal, though more often it is simply a consequence of having many lines. A simple test: if removing all grids but one destroys your reason for the trade, there was no reason.
What can genuinely be measured here
Not "does 0.618 work" but how retracement depths are actually distributed for YOUR instrument. That is computed on your own history and gives a real answer: where corrections end more often and where almost never. It may turn out that your coin's pullbacks most often fit into 30–45%, in which case the 0.618 line is a rare case for it rather than an expectation. This knowledge cannot be bought in a course: it is different for every instrument.
A Fibonacci level has a test it usually lacks: look at the density. If large limit orders are sitting at the computed price right now, the level stops being geometry and becomes a place where other people's money is. If they are not, you are looking at a line you drew yourself. Same caveat as always: the book is alive and a wall can be pulled a second before price arrives.
Order-book densityThe distribution of pullback depth, computed by you
Take one instrument and find thirty completed corrections after clear impulses. For each, compute the depth of the pullback as a percentage of the impulse and sort them into bins: under 30%, 30–45%, 45–60%, 60–80%, over 80%. Chart the bars. You will get a distribution found in no course, because it is yours. And you will also see how many corrections went deeper than 80% — that is, how many times a "deep pullback" actually meant the impulse was labelled wrongly.
Price bounced three times exactly off 0.618. Does that prove the level works?
Not on its own. First check how many other lines were on the screen: with three grids a bounce "exactly off a level" happens almost always, because there are many levels. Then shift the anchors by one candle and see whether the coincidences survive. If the picture fell apart, you were looking at fitting. A robust observation survives a shift to the neighbouring anchor; a fitted one does not.
If there is no mechanism, why do the levels sometimes work anyway?
For the same reason the neckline of a head and shoulders works: many people draw them, and orders and stops genuinely accumulate at the computed prices. That is an honest mechanism, but it has a limit: it weakens the less known the instrument is and the fewer people look at that scale. On bitcoin, daily Fibonacci levels are seen by half the world; on a coin three weeks old, by roughly nobody — and there the line stays geometry.