8 minute read · Ultimo Research Desk · Reviewed 5 Sept 2026
Fibonacci Retracements: What They Measure and When They Lie
A Fibonacci retracement takes a price swing you chose — a low and a high — and draws horizontal lines at fixed fractions of it: 23.6%, 38.2%, 50%, 61.8%, 78.6% of the way back. Traders watch for a pullback to pause at one of those lines. The lines are exact once the swing is chosen, and that "once" is the entire story.
They lie because the swing is chosen by a person, after the move, and there are always several swings to choose from. A grid that "worked" usually worked because the analyst picked the anchors that made it work. The ratios contain no time, no volume and no knowledge of why the market moved; they are fractions of a distance.
What they measure
The ratios come from the Fibonacci sequence, each term the sum of the previous two: successive ratios converge on 0.618, and 0.382 and 0.236 are its powers. The sequence is real mathematics, described in Leonardo of Pisa's Liber Abaci in 1202 — a book about arithmetic and bookkeeping that says nothing about price charts. The application to markets is a twentieth-century charting convention, popular enough to be self-fulfilling at times, and that is the strongest case for it.
Two of the standard levels are not Fibonacci at all: 50% is a midpoint, and 78.6% is the square root of 0.618, added by chartists later. Nobody should be told otherwise.
The formula, in words
For an upswing from low L to high H, each retracement level is H − ratio × (H − L). For a downswing, L + ratio × (H − L). Extensions — 127.2%, 161.8% — project beyond the swing and are a separate tool with the same caveats.
Worked example
A move from 100 to 140. Range 40.
- 23.6%: 140 − 9.44 = 130.56
- 38.2%: 140 − 15.28 = 124.72
- 50%: 140 − 20 = 120
- 61.8%: 140 − 24.72 = 115.28
- 78.6%: 140 − 31.44 = 108.56
Change the anchor low from 100 to 96 — a perfectly defensible choice if the swing started with a wick — and every level moves: the 61.8% goes to 112.81. Nothing about the market changed; the grid did.
How they are read
A level is treated as a zone to watch rather than a price to trade: a pause, a rejection candle, a momentum turn near it. Several grids from different swings sometimes put levels close together, and the cluster gets called "confluence" — which is either a real crowd of orders or the same data drawn three times, and the chart cannot tell you which.
The reproducible way to use them is to fix the swing rule first — the last confirmed pivot on a stated timeframe, say — and never adjust it after seeing where price went.
When they lie
Anchors chosen with hindsight; that is the main one. A level "held" because one candle touched it once, when ordinary noise touches many lines. In a fast market, price passes three levels before the bar closes, and it did so because information arrived, not because 38.2% "failed". Different timeframes give different swings and contradictory grids on the same screen.
And no time: a 38.2% retracement can arrive in one bar or over two months, and the line says nothing about whether the pullback is a pause in a trend, the start of a new one, or a reaction to one headline.
What they do not tell you
Value, the probability of a bounce, when a level will be reached, which way a break goes, why the trend began, volume, volatility, liquidity — or which high and low are the "correct" anchors. That is the one question the tool exists to answer and the one it leaves entirely to you.
What the evidence actually says
Three papers come up in every serious discussion of indicators, so it is worth knowing what they found rather than what people say they found. Brock, Lakonishok and LeBaron (1992) tested simple moving-average and trading-range rules on ninety years of the Dow and found they carried information relative to a random benchmark. Sullivan, Timmermann and White (1999) then re-ran that idea across nearly eight thousand rule variants and showed that once you account for how many rules were tried, the best-looking one is far less impressive — the "data-snooping" result. Park and Irwin (2007) reviewed ninety-five later studies and found roughly half positive, a quarter negative and the rest mixed, with results weakening after transaction costs and risk adjustment.
The fair summary is not "indicators work" and not "indicators are astrology". It is that a fully specified rule — entry, exit, size, costs — can be tested, and most rules that look good on a chart do not survive the test. Whatever you build on this indicator, test it as a complete rule on data it has not seen.
There is no direct peer-reviewed evidence that the named ratios have predictive value across markets. Fibonacci analysis has more researcher degrees of freedom than almost any other tool — anchors, timeframe, ratio set, confirmation, exit — which is precisely the situation the data-snooping result describes. If you want to test a grid, decide the anchor rule in writing first.
Where it fits
Not a vote on our signal pages, because there is no objective swing for a program to choose. If you draw one on MT5, mark the anchors you used and the timeframe, so that next week you can tell whether the level held or whether you moved it. Pivot points are the mechanical alternative: no anchors to choose, and the same levels for everyone.
Risk line: the level everybody can see is watched, and watched levels sometimes hold. Sometimes is not a plan; the stop has to be somewhere the idea is actually wrong.
Sources
- Leonardo of Pisa (Fibonacci), Liber Abaci (1202), digitised edition: Museo Galileo
- Murphy, John J., Technical Analysis of the Financial Markets (1999): Google Books
- Park and Irwin (2007): DOI
- Sullivan, Timmermann and White (1999): DOI
- Lo, Mamaysky and Wang (2000): DOI
- Brock, Lakonishok and LeBaron (1992): DOI
Written by the Ultimo Research Desk and checked against our own contract specifications and client agreement before publication; reviewed again when those change. Educational only — nothing here is a recommendation to trade. Spotted an error? Tell us.
