Skip to main content
Global Market Radar

How we read a chart

Fibonacci retracement

Five subdivisions of one chosen swing — and the reason the two anchors matter more than the five numbers.

Assumes:
The basics of this family
Markets:
Crypto · Equities · Indices
Timeframes:
15m · 1h · 2h · 4h · 1d

What it means

A retracement grid takes one completed move — a swing low up to a swing high, or the reverse — and marks how far back through it price has come. The conventional marks are 23.6%, 38.2%, 50%, 61.8% and 78.6%. Only three of those are Fibonacci ratios; 50% is simply the halfway point, included by convention. This publication includes it too, because every charting package does and a reader comparing four levels against their own five would reasonably assume our engine was broken.

The grid is ARITHMETIC OVER TWO POINTS. That sentence is the whole concept. Nothing about the ratios obliges a market to turn at them, and this publication has no evidence that they are respected more often than any other subdivision of the same distance. What the grid gives a reader is a shared vocabulary for where inside a move price currently sits.

Which is why the two ANCHORS are the real content. The same instrument on the same timeframe produces an entirely different set of five numbers from a different swing pair — and almost every retracement level quoted elsewhere omits the anchors it was drawn between, which is exactly what makes it impossible to reproduce.

How we detect it

The engine needs at least thirty bars. It finds swing highs and lows with the same fractal rule the zones use: a bar is a swing only once two bars have closed on both sides of it. That is what CONFIRMED means here, and it is why the newest bars are never anchors — a high is not a swing high until the market has failed to beat it. The grid therefore does not redraw itself from under a reader on the next tick.

The pair chosen is the newest confirmed swing and the nearest preceding swing of the opposite kind: the leg the market has most recently completed. Both anchor prices and both anchor times travel with the grid and are printed beside it, so any reading can be checked against the reader's own chart.

Where an EARLIER opposite swing would produce a leg of comparable size — within 15% of the chosen one — the engine reports that as an alternate rather than presenting either as the answer. A much larger earlier leg is not treated as an alternate: that is a bigger swing, a different degree of move, and a question about a different timeframe.

A leg spanning less than one Average True Range is refused outright. The arithmetic still works; its five results sit inside a single ordinary bar and cannot be told apart from noise.

If the reader supplies two prices, the engine measures between those and labels the result as theirs. It does the arithmetic and none of the choosing.

Where support and resistance zones have also been drawn, the engine reports which levels fall inside one — but only for zones that were NOT built from the grid's own anchor swing. A zone containing an anchor is the same observation counted twice, and reporting it as agreement between two methods would be a confirmation this engine manufactured rather than found.

When the read carries information

  • The move being subdivided is one a reader would also identify as a move: a clear leg with a beginning and an end, rather than a stretch of chop the swing detector happened to find two points in.
  • The anchors are visible. A grid is only as good as the pair under it, which is why this publication prints them and treats a grid without them as unpublishable.
  • Something drawn independently agrees. A level that also sits inside a zone built from repeated touches is being pointed at by two different readings of the chart; a level that sits alone is one subdivision of one chosen distance.

When it misleads

  • The leg is ambiguous. On most charts more than one swing pair is defensible and the five numbers move substantially between them. This is the ordinary case rather than the exception, and it is why the engine reports alternates instead of hiding the choice.
  • The market has repriced. After a structural event the earlier swing describes a market that no longer exists, but it is still on the chart and the detector still finds it.
  • The reading is treated as a forecast. The grid says where price is inside a move that already happened. It says nothing about the next one, and this publication never presents it as though it did.
  • The series is stale. A grid is a statement about where a market IS, so a feed that has stopped produces a refusal here rather than a well-formed answer about a dead chart.

How it reaches you

On our surfaces a grid is always printed with its anchors — the timeframe, both prices, both dates and the direction of the leg — then the five levels, then the line saying it is a reference and not a rule the market follows.

Where no grid can be drawn the reason is named: too few bars, no confirmed pair, a leg too small to subdivide, or a series too old to read as current. Those are four different things and a reader is told which.

Nothing here is a prompt to act. A retracement level is context for reading a chart, and this publication does not tell anyone what to do about one.

Worked example

No worked example has been planned for this concept.

This page stays text only until every item above is settled. Nothing is loading, and there is nothing here to open.

Ask a market question

Market-reference material for understanding our coverage. Global Market Radar does not issue trading signals and does not give investment advice.