Fibonacci Retracement: How to Draw and Read It

Fibonacci retracement has a reputation it does not quite earn. The ratios come from the Fibonacci sequence, which appears in sunflower heads and nautilus shells, and that lends the tool an air of natural law. In practice it is a ruler: you pick a swing low and a swing high, and it marks off fixed fractions of the distance between them. Used that way, as a set of reference lines rather than a prophecy, it is genuinely handy.

Where the numbers come from

In the sequence 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89… each number divided by the next approaches 0.618. Divided by the number two places ahead it approaches 0.382, and three places ahead 0.236. Charting tools add 50%, which is not a Fibonacci ratio at all but a round halfway mark traders watch anyway, and 78.6%, the square root of 0.618.

For a pullback in an uptrend, each level sits at: level = swing high − (swing high − swing low) × ratio.

Drawing it on a real swing

Apple bottomed at an intraday low of 169.21 on April 8, 2025, during the tariff sell-off, and rallied to an intraday high of 288.62 on December 3, 2025. The swing is 288.62 − 169.21 = 119.41.

Case 1: a pullback that respected the level

After the December high Apple drifted lower for seven weeks. On January 20, 2026 it traded down to 243.42, a retracement of 37.9% and just 0.41 above the 38.2% level, then closed at 246.70. Within three weeks it bounced to 280.91 on February 6, 15.4% off the low. It chopped lower again through March, but it never traded below 243.01 before setting a new high of 292.13 on May 7.

That is close to a textbook picture. A shallow pullback to 38.2% in a strong uptrend is what the tool's fans point to. It is worth being precise about what the level did and did not do, though. It did not predict the date of the low, or that the bounce would happen at 243 rather than 250 or 235. What it offered was a pre-computed price where a trader could ask, "if buyers are still in charge, is this where they show up?" and a clear line below which that idea would be wrong.

Change the anchors, change the answer

Draw the same swing on closing prices instead of intraday extremes and the picture shifts. Apple's lowest close was 172.42 on April 8 and its highest 286.19 on December 2, which puts the 38.2% level at 242.73. The deepest close of the pullback was 246.63 on March 30, 2026, a 34.8% retracement that sits between levels rather than on one. Same stock, same months, two defensible drawings, two different stories. Before trusting any Fibonacci "hit", check which anchors produced it.

Case 2: a pullback where every level failed

Alphabet rallied from an intraday low of 147.22 on September 9, 2024 to 207.05 on February 4, 2025. Its retracement levels were 177.13 (50%), 170.08 (61.8%) and 160.02 (78.6%). Here is how long each lasted:

The stock kept falling to an intraday low of 140.53 on April 7, a 111% retracement. The 61.8% level, the so-called golden ratio, gave way two sessions after the 50% level did. Anyone who bought at 61.8% because it is "the" Fibonacci level, without a plan for being wrong, rode it through two more levels and past the start of the swing.

Why levels look more accurate than they are

Five levels cut a swing into slices only 12–17% of its size. On Apple's swing, neighbouring levels were 14 to 20 dollars apart, so most pullbacks end within about 10 dollars of one line or another. With enough lines, something always looks like it held. Two habits keep the tool honest:

• Decide which level you care about before price gets there, not after.
• Give more weight to a level that lines up with something else: a prior support or resistance zone, a rising moving average, or a gap.

A plan built on a level

Fibonacci is most useful for turning a vague idea ("buy the dip") into numbers. Using Apple's January levels as an illustration, not a recommendation: a trader who wanted to buy a pullback might watch 243.01, place the invalidation under the 50% level at 228.92, and compare that risk with a first target at the old high of 288.62. From 243.01, that is 14.09 of risk against 45.61 of potential, about 1 to 3.2. The level gives the plan its shape; position sizing decides how much the plan can cost.

To build a feel for it, draw retracements on a past swing while the following candles are hidden, pick your level, and then reveal what happened. You will quickly learn how often the "right" level is only obvious afterwards.