Pull up any gold chart after a strong rally and you will find someone who drew a retracement from the swing low to the swing high, watched price stall near 61.8 percent, and concluded the ratio caused it. Pull up the same chart with the anchor moved two candles earlier and the level that price respected is now 50 percent. Both traders will tell you Fibonacci works.
The tool is genuinely useful. It is useful for a narrower reason than most courses claim, and the difference between the two readings decides whether it improves your entries or just decorates the chart.
Where the ratios come from
The Fibonacci sequence adds each number to the one before it: 1, 1, 2, 3, 5, 8, 13, 21. Divide any term by the next and the result converges on 0.618. Divide by the term two places along and you approach 0.382. Skip three places and you get roughly 0.236. Those three ratios, plus the square root of 0.618 at about 0.786, are the retracement levels every charting package draws.
Then there is 50 percent, which is not a Fibonacci ratio at all. It comes from Dow theory, where a correction of about half of a prior advance was considered ordinary. It stayed on the tool because traders watch it, which is a circular reason and also the only reason any of these levels matter.
| Level | Origin | Typical use |
|---|---|---|
| 23.6% | Fibonacci ratio | Shallow pullback in a fast trend |
| 38.2% | Fibonacci ratio | First serious pullback zone |
| 50% | Dow theory convention | Watched by habit, often the busiest |
| 61.8% | Golden ratio | Deep pullback, trend still intact |
| 78.6% | Square root of 0.618 | Last-chance level before the move is void |
| 127.2% / 161.8% | Extensions | Target projection beyond the swing |
Anchoring, which is the whole game
A retracement is defined by two points. Choose different points and you get different levels, so the discipline lives entirely in how those points are selected. The rule I use is simple: anchor to the swing that produced the structural change, meaning the low that started the impulse and the high that broke the previous high. If a move did not break anything, it is noise, and there is nothing to retrace.
Two secondary decisions follow. Wicks or bodies: pick one and never change it mid-analysis. Which timeframe: the retracement of a daily impulse is a different object from the retracement of a five minute one, and mixing them produces a chart with nine levels where every price is near something. Our note on chart timeframes covers keeping those layers separate.
If you have redrawn the tool because price ignored the first level, you are no longer analysing. You are curve-fitting a picture to an outcome you already saw. Draw once, and if the level fails, record it as a failure.
Confluence beats the ratio
A retracement level on its own is a coin flip with a nice number attached. What changes the odds is what else sits at the same price. A 61.8 level that lands on a prior swing high that has flipped to support, on the round number, and inside the zone where the impulse originated, is a place where several groups of participants have a reason to act. A 61.8 level in the middle of empty space is a line on a screen.
This is why I treat Fibonacci as a filter rather than a signal. The sequence is: decide the trend context first, mark the structural levels that already exist, then draw the retracement and see whether it agrees. If it agrees, the zone gets tighter and the stop can sit just beyond the structure. If it disagrees, the structure wins. Traders working from price action rather than indicators tend to arrive at the same ordering.
Extensions and where targets actually come from
The extension levels, 127.2 and 161.8 most commonly, project beyond the end of the measured swing and are used as targets. They have the same status as the retracements: a place where a cluster of take-profit orders may sit, with no mechanism guaranteeing price gets there.
The practical use is in ratio planning rather than prediction. If the entry sits at a 50 percent retracement, the stop below the swing low and the nearest extension gives roughly three times the risk, the trade clears a sensible risk to reward threshold before you take it. If the same geometry only offers one times risk, the setup is not worth the spread and the swap, whatever the level says.
The abuse cases
Four habits do most of the damage. Drawing on a range, where there is no impulse to retrace and every level is arbitrary. Stacking retracements from four different swings until the chart is a grid. Moving the anchor after entry so the losing trade looks like it is still inside a valid zone. Treating a level as an entry trigger by itself, with no confirmation from how price behaves when it arrives.
The last one deserves the most attention because it is the most seductive. A limit order resting exactly at 61.8 will be filled in every trend that ends there and every trend that keeps going. Waiting for a reaction at the level, a rejection candle, a failure to make a new low, a shift in the smaller timeframe structure, costs a few pips of entry and removes a whole category of losses. That is the same argument made in trend following basics, and it applies here without modification.
Trading leveraged instruments carries a high risk of loss, and no drawing tool changes that. Fibonacci is a way of organising where you look. Position size and stop placement decide what happens when you look in the wrong place, which is why they belong to the plan rather than to the chart.
"If your fib lines up with a level that was already there, it tells you something. If it lines up with nothing, you have drawn a decoration and paid for it with a stop."
— Alex Onta, Executive Director, SINGUARD
Key Takeaways
- The levels matter because orders cluster near them, not because of any property of the sequence itself.
- Anchor only to swings that broke structure, use wicks or bodies consistently, and keep timeframes separate.
- A retracement is worth trading when it agrees with a level that already existed on the chart.
- Redrawing the tool after price ignores a level converts analysis into after-the-fact justification.
Frequently Asked Questions
Why is 50 percent included when it is not a Fibonacci number?
It is a convention borrowed from Dow theory, where a correction of roughly half a prior move was treated as normal. Charting packages kept it because traders watch it. Its presence on the tool is habit rather than mathematics.
Should I draw Fibonacci from the wick or the candle body?
Pick one and apply it to every chart. Wicks are the more common choice because they mark the actual extreme of the move. What matters far more than the choice is that you do not switch method after the fact to make a level line up with price.
Do Fibonacci levels work because the market respects them?
There is no mechanism forcing price to react at a ratio. The levels attract orders because a large number of traders and some automated systems place them there, so the effect that exists is reflexive. Treat a level as a place to look for evidence, not as a reason on its own.