Fibonacci Extension: A Target, Not a Level
A Fibonacci extension projects the ratios beyond the end of a completed move to mark possible targets, using three points rather than two. It answers where a move might reach, which is a weaker and more speculative claim than where a pullback might stop.
How it works
A retracement grid divides a move; an extension continues it. The retracement asks how far back price might come. The extension asks how far forward it might go once the pullback ends.
That extra question costs an extra point. A Fibonacci retracement needs a swing low and a swing high. An extension needs those two plus the end of the pullback — and the pullback is only finished in hindsight, so the third point is provisional every time you place it.
The common levels are 1.272, 1.618 and 2.618. They are the square root of 1.618, the ratio itself, and its square. The arithmetic is exact; what the arithmetic is applied to is not.
Target and level are different claims
A support level is a place where resting interest might exist. A target above every price ever traded in the move has no resting interest behind it at all — nobody has a position there yet.
On this site’s shared history the second leg ran anywhere from half the first to more than twice it. The tenth percentile was 0.512 and the ninetieth was 2.195, across 82 measured pairs.
A four-fold spread is not a target. Placing an exit at 1.618 inside that distribution is picking one number from a wide range and calling it a plan.
The median second leg was 0.958 of the first — almost exactly the size the simple projection predicts, and a number that describes the middle of the distribution rather than any particular trade.
In practice
Watch participation into the target rather than the target itself. Falling volume as price approaches an extension is information; the line arriving is not.
Match the extension’s timeframe to the holding period. A daily 1.618 is a multi-week target, which is not a level an intraday position can wait for.
A gap can jump two levels at once, which is why a limit exit sitting at 1.618 can be filled at a far better price or missed entirely.
The tool says nothing about risk. A target is half a trade; the stop comes from structure, and the position size comes from the distance between them.
Scaling out across three extensions pays three round trips. Each one is 2% of a median bar’s range on this history, which is a real cost attached to a purely cosmetic exit schedule.
The one honest argument for the levels is other people. Enough traders place exits at the same ratios that the orders cluster, which is a self-fulfilling mechanism rather than a mathematical one — and it works only where the swing everybody measured is obvious enough that they all chose the same two points.
There is a practical distinction between the near and far levels worth making. The 1.272 sits close enough that a move of ordinary size reaches it often; the 2.618 requires the kind of leg that happens a few times a year. Treating them as a single family of targets hides that difference, and it is the difference that decides whether an exit ever fills.
Which suggests one defensible way to use the tool. Set the first exit at the near level, where the hit rate is high and the profit small, and leave a remainder running with a trailing stop rather than a distant line. That converts the far extension from a target into an upper bound — a number you would be pleased to reach rather than one the plan depends on reaching.
What a Fibonacci extension is not
It is not a forecast. It marks a possibility with no probability attached.
It is not a retracement. Different question, different number of points.
It is not an entry signal. Nothing about it says the move will start.
And it is not a substitute for a stop. It only describes the profitable side.
When it fails
In a range the first extension usually sits at the opposite boundary, which makes the target look sophisticated while it is really just the top of the box.
The second failure is choosing the third point after the fact. Once the move has run, one pullback low produces a target price already reached and another does not, and the temptation is to pick the flattering one.
A third is the level that was never reachable. A 2.618 extension on a five-minute swing can be further away than the instrument moves in a session.
A fourth is holding through an obvious exit. The most expensive use of the tool is as a reason to ignore what price is doing in front of you.
And a fifth is treating the median as the plan. A distribution running 0.51 to 2.20 has a tidy middle and no reliability anywhere.
The original data
On this site’s shared 576-bar history, 82 consecutive swing pairs were measured on a one per cent zigzag.
The median second leg was 0.958 of the first, the tenth percentile 0.512 and the ninetieth 2.195, and only
16% landed within a tenth of equality. The figures are in research/series-measurements.json, produced by
site/measure_series.py.
Two cautions about that measurement. The series is generated, not a real market, so it contains no Fibonacci behaviour by construction — which makes it a useful control rather than evidence about any instrument. And 82 pairs is a small sample. What it does show cleanly is the shape of the problem: a projection can be correct in the middle of its distribution and still be wrong on five trades out of six. The check worth running on your own market is the same one: measure the second leg against the first across a hundred swings, and see how wide your distribution is before you place an exit on one line in it.
Related
Fibonacci covers the retracement grid this extends. Measured move is the same projection without the ratios. And take profit is where an exit decision actually belongs.
I use extensions the way I use a weather forecast for next Thursday. It tells me roughly what to plan for and I do not commit anything to it. The trades where I have been hurt by them are the ones where I held past an obvious exit because a line further up said the move had further to go.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.