Fibonacci Retracements
Horizontal levels drawn at conventional fractions of a prior move. Their mathematical justification is weak and their practical relevance comes from how many people draw them.
MadStockAlerts Research · Updated August 28, 2026
What to take away
- Levels are drawn at 23.6, 38.2, 50, 61.8 and 78.6 percent of a prior move.
- The 50 percent level is not a Fibonacci ratio at all; it is included by convention.
- The choice of swing high and low is subjective, so two analysts get different levels.
- Their usefulness is largely self-fulfilling, which does not make it unreal.
- They are best used on one obvious swing rather than on every candidate.
MAD Academy Training Video · 0:45
Why the Levels Work When They Work
The numbers have no special property. What they have is a very large number of people drawing the same lines.
This lesson is part of a Stock Alerts + Tools plan.
The construction
Pick a swing low and a swing high, and horizontal lines are placed at set percentages of the distance between them. The premise is that a pullback within a trend often stalls at one of these fractions before the trend resumes.
| Level | Origin |
|---|---|
| 23.6% | Derived from the Fibonacci sequence |
| 38.2% | Derived from the Fibonacci sequence |
| 50% | Not a Fibonacci ratio. Included because half a move is a natural reference |
| 61.8% | The golden ratio; the most watched level |
| 78.6% | The square root of 61.8 percent |
The presence of 50 percent in a list of Fibonacci ratios is worth noticing. It is there because traders watch it, which is precisely the argument that would justify the whole toolset without needing the mathematics at all.
Scroll the chart sideways to see all of it.
The honest assessment
There is no accepted mechanism by which a sequence describing rabbit populations and sunflower seed heads should govern the price of equities. The literature testing these levels against randomly chosen ones does not produce a clear edge.
The defensible argument is behavioural rather than mathematical. Enough participants draw the same levels from the same obvious swings that orders cluster near them, and clustered orders are real supply and demand. That is a claim about crowd behaviour, not about number theory.
Accepting that framing has a useful consequence: the levels are worth more on liquid, widely followed securities where many participants are drawing them, and worth much less on an obscure name that few people are charting at all.
The subjectivity problem
The levels depend entirely on which swing high and low are chosen, and on any chart there are several plausible candidates. Two analysts working the same chart routinely produce different level sets, and after the fact it is always possible to find a pair of points that makes a level look respected.
The practical mitigation is to use only the most obvious and most recent swing, to draw the levels before the pullback rather than during it, and to treat a level as one reference among several rather than as a line the market is obliged to honour.
A related discipline: if a chart needs three different Fibonacci overlays to explain its behaviour, the overlays are not explaining anything. Enough lines will always intersect something.
Where the numbers come from
The ratios used are derived from the Fibonacci sequence, where each term is the sum of the two before it. As the sequence extends, the ratio of consecutive terms converges on approximately 0.618, and the other levels are simple derivations from it.
| Level | Where it comes from |
|---|---|
| 23.6% | The ratio of a term to the term three places later |
| 38.2% | The ratio of a term to the term two places later |
| 50% | Not a Fibonacci ratio at all. It is simply half |
| 61.8% | The golden ratio, the sequence's limiting ratio |
| 78.6% | The square root of 61.8 percent |
The middle row is worth noting because it is the most watched level of the set and has nothing to do with the sequence. It is included by convention, and its presence is a reasonable indication that the set is used for practical reasons rather than mathematical ones.
There is no accepted mechanism connecting this sequence to prices. The defensible reading is that the levels are watched by enough participants to occasionally become self-fulfilling, which is a claim about coordination rather than about mathematics.
The subjectivity that cannot be removed
A retracement is drawn between two points, and both are chosen. That makes the resulting levels a function of a decision rather than of the data, and the decision is made after the move it describes has happened.
- Which swing high and low: a chart contains many, at every timeframe, and each pair produces a different grid.
- Wicks or bodies: using the extreme prints or the closes moves every level.
- Which timeframe: the same security has a daily grid, a weekly grid and an hourly grid, all valid.
- Whether extensions beyond 100 percent are included, which adds levels above the move entirely.
The combined effect is that a chart with several retracement grids has a level near almost any price. Once that is true, price bouncing near a level is not evidence, because there was no price at which it could have bounced without being near one.
The discipline that makes the tool defensible is the same one that makes any drawing defensible: one grid, drawn from a stated swing, before the retracement happened, and left alone afterwards.