What Is Fibonacci Retracement?
Fibonacci retracement takes a swing from a low to a high and marks fixed fractions of that distance — 23.6%, 38.2%, 50%, 61.8% and 78.6% — as potential support during a pullback. The arithmetic is exact, but the swing high and low you measure from are chosen by you.
Covered on this page: TradingView.
The tool with the most impressive-sounding name on the chart, doing something completely ordinary: dividing one distance into fractions.
How it works
Pick a swing low and a swing high. In the worked example, 99.70 and 101.47.
Then mark fixed fractions of that distance below the high:
23.6% = 101.05 · 38.2% = 100.79 · 50% = 100.58 · 61.8% = 100.38 · 78.6% = 100.08
That is all the arithmetic there is. The distance is 1.77, and each level is the high minus a fraction of it.
50% is not a Fibonacci number
Worth knowing because it is on every version of the tool and nobody mentions it.
61.8% comes from the Fibonacci sequence — divide any term by the next and it converges on 0.618. 38.2% is 0.618 squared. 23.6% is 0.618 cubed.
50% is halfway. It has no connection to Fibonacci at all. It is there because traders watched the midpoint of a move long before anyone put a sequence on a chart.
And it is the level price actually reacted to below. Which is either an argument that the sequence is not doing the work, or an argument that it does not matter — both readings are available and this page will not pretend to settle it.
What actually happened here
The pullback turned at 100.54. The nearest level is 50% at 100.58 — 0.05 away, about 3% of the swing.
Then price continued to a high of 101.79, past the old one. Textbook.
Note what that sentence hides. “The nearest level” was identified after the low was in. At the moment price was falling through 100.79, there were three more levels below it and nothing said which one would hold.
The golden pocket
In practice most people ignore three of the five levels and watch the band between 50% and 61.8%.
That narrows the tool to a zone rather than a line, which is more honest — price reacts to areas, and expecting a turn at an exact price is expecting more precision than any of this has.
The real weakness
Everything above depends on two points you chose.
In that example there is a second, perfectly defensible low. Measured from it, the 61.8% level moves from 100.38 to 100.49 — a different price, from the same tool, on the same chart, with no mistake made by anyone.
This is the whole argument about Fibonacci and it does not resolve. The defense is procedural: draw it before you have a position, use the most obvious swing on the chart, and write down which two points you used. If you find yourself re-picking the swing, you are not measuring, you are negotiating — the same failure as the trend lines page.
The other half: extensions
The same tool run past the high, used for targets rather than for support.
Extensions project the swing beyond its own high: low + (high − low) × ratio, with 127.2% and
161.8% the two people watch. In the same example that is 101.95 and 102.56.
Price reached 101.79 and stopped. Short of both — which is the ordinary outcome and worth seeing once, because extension targets are usually presented alongside the moves that reached them.
They are weaker than retracements, and for a specific reason. A retracement level sits inside territory price has already traded, so there are real orders and real memory there. An extension sits in a range price has never visited, where the only thing at that price is arithmetic.
A worked example
Find the most obvious impulse on the chart — the one a stranger would pick.
Draw from its low to its high, and leave it alone.
Watch the 50–61.8 band. Between 100.58 and 100.38 here.
Wait for something else to agree. A level price already reacted at, a swing low, an old high. A Fibonacci level with nothing else at it is a line you drew.
The invalidation is a close below the swing low — 99.70 — because past that the swing you measured from has stopped existing.
The original data
Across our study of 24,971 trading videos, 381 have Fibonacci or fib in the title. The median one gets 8,105 views, 68% never pass 50,000, and the median length is 10.0 minutes.
381 is one of the larger fields measured for this glossary — more than chart patterns at 287 and two and a half times Ichimoku at 151.
The corpus carries description text for 85 of those 381, and across those 85, 4 mention invalidation, failure, or what a bad read looks like.
When it fails
Price goes through every level
All five levels, and then the swing low itself.
A retracement that passes 78.6% is barely a retracement any more, and one that passes the low has ended the move you were measuring. There is no sixth level to fall back on — the tool is finished, and holding on for one is inventing a level.
There is always a level nearby
Five lines across one swing means no part of the pullback is more than about 12% of the swing away from one of them. So “it reacted at a Fibonacci level” is very hard to say after the fact and be wrong.
The test that survives: did you name the level before, and did anything other than the line agree with it?
The swing was chosen to fit
Covered above and it is the real one. A different low gave a different 61.8. If the level moves whenever the trade needs it to, nothing is being measured.
You found the level once price had turned
With the ending covered, three levels sit below price and all of them look plausible. Afterwards exactly one of them is the obvious answer, and that difference is the whole difficulty.
Related
Pullback is what these levels are measuring, and when one stops being a pullback at all.
Support and resistance is what turns a drawn line into a real one — the confluence test above.
And swing highs and lows decides which two points you measure from, which is the choice everything here rests on.
Much of what makes these levels work is that thousands of other people drew the same two points on the same obvious swing. That is not a criticism; it is the mechanism. Do not treat a level as a reason on its own: a 61.8 that lines up with a level price already reacted at is worth something, and one floating in the middle of nowhere is just a line.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money. Some links on this page earn a commission if you buy through them. It costs you nothing and it does not decide what appears here or in what order — how these pages are made is set out in our methodology.