How to Draw Fibonacci Retracements: 0.382, 0.5 and 0.618

What retracement levels are

After a big move in one direction, price often gives back part of it. Fibonacci retracement is a way to estimate where that pullback might stop: you divide the size of the previous move (the swing) by fixed ratios and draw horizontal lines. The lines themselves are simple arithmetic, so anyone who picks the same two points gets the same values. What differs is which high and low you pick as the swing, and that is also the weakness this article spends the most time on. Pullbacks are explained by how market participants behave. After a big rise, selling comes in from people locking in profits, late arrivals wait for a slightly lower price, and those who missed the move vow to buy if it comes back down. Where these orders will meet is not fixed, so people have used lines that divide the move by set ratios as reference points. It is more accurate to see Fibonacci retracement as a tool that lets everyone calculate those reference points the same way. Thinking that the lines pull price toward them is a common mistake.

How to draw them: from the low to the high

For an upswing, draw from the low to the high; each level is high − (high − low) × ratio. With a low of 100 and a high of 150, the swing is 50 and the levels are as below. 0% is the high and 100% is the low itself; a pullback all the way to 100% has retraced the entire swing. For a downswing, only the direction is reversed: draw from the high to the low and calculate low + swing × ratio. This site's Fibonacci tool automatically takes the highest high and lowest low in the chosen range as the swing, and tells upswings from downswings by which of the two came first. When drawing by hand, the order of checks is this. First choose the bar length, and pick one clearly visible large move on that bar length. Next decide whether the start and end points are wick tips or bodies, and do not change that method on the same chart. Finally, write down the level values and compare how far actual price came down, as with the 61.8% level in the figure. A common mistake is drawing it upside down so that 0% and 100% swap places, so always check that 0% sits at the end of the swing.

• 23.6%: 150 − 50 × 0.236 = 138.2
• 38.2%: 130.9
• 50%: 125
• 61.8%: 119.1
• 78.6%: 110.7

Where the ratios come from

The Fibonacci sequence, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55…, makes each number by adding the two before it. As the numbers grow, one number divided by the next approaches 0.618, divided by the number two places later it approaches 0.382, and divided by the number three places later it approaches 0.236. For example, 34 ÷ 55 ≈ 0.618, 21 ÷ 55 ≈ 0.382 and 13 ÷ 55 ≈ 0.236. 78.6% is the square root of 0.618 (about 0.786); the 161.8% extension level is the golden ratio, 1.618, and 127.2% is the square root of 1.618. This site's tool calculates extensions as the prices 127.2, 161.8 and 261.8% of the swing away from the swing's starting point, but the reference point for measuring extensions can differ between tools. The idea that these ratios show up often in nature is widely known, but that is no reason market prices should stop at them. Having a mathematically tidy origin and price reacting at that spot are separate questions. So the origin of the ratios only explains why these particular numbers came into use; whether the levels are useful has to be judged separately. When matching values with another tool, first check that the extension reference point and the list of levels are the same.

50% is not a Fibonacci ratio

The 50% level that always appears in the list is not a ratio from the Fibonacci sequence. The ratios from the sequence settle on values such as 0.618, 0.382 and 0.236 as the numbers grow, and 0.5 is not among them. 50% is commonly explained as having joined the list from a convention of treating a retracement of about half a move as important. Values like 78.6%, which put a sequence ratio through one more calculation, are mixed in as well, so the list of levels itself differs slightly between tools and people. Having Fibonacci in the name does not mean every line rests on the same basis. Still, 50% never drops off the list because people really do watch that price a lot. 'It has retraced half' is easy to understand and shows up often in news and market commentary. Many take the view that whether a level draws attention because of the sequence or because of convention, the fact that many people watch the same price matters more. A common mistake is treating 61.8% as 'a stronger level because it is the golden ratio'; nothing in this calculation ranks one level as stronger than another. Some tools drop 23.6% or 78.6% or add other values, so check the list in the tool you use first.

The biggest weakness: which swing do you pick?

Every level comes from the swing's two points, so changing just one point moves all the levels. If the swing in the first figure starts from the intermediate low B (116) instead of the lowest low A (100), the same bounce is no longer a bounce off the 61.8% level but a move that has retraced almost the entire swing. Whether you use wick tips or bodies, and whether you pick the swing on the daily or the 4-hour chart, also changes the values. The thing to be most careful about is that if you pick a suitable swing after the bounce is over, any bounce can be explained by one of the levels. One way to reduce this problem is to set the rule for picking the swing before the bounce happens. If you define it mechanically, for example as 'the highest high and lowest low within the last so many bars', it becomes hard to fit it after the fact. That is also why this site's tool picks the highest high and lowest low in the range automatically. But how you set the range raises the same problem again, so it helps to change the range and see how much the levels move. Drawing levels on smaller swings inside a big swing fills the chart with overlapping lines, so looking at one swing at a time is less confusing.

Prices where other evidence overlaps

The Fibonacci levels that draw attention are usually prices where other evidence overlaps. When a previous high or low, a support or resistance zone where price has stopped several times, a moving average or a price zone with heavy traded volume sits near a level, more people are likely to be watching that price. In the first figure, too, the 61.8% level (119.1) lies within the price zone where the rally paused for a while (about 116 to 121.7). But when price stops at such a spot, you cannot tell whether it stopped because of the ratio or because of the other evidence that overlapped it. Looking at the price zones where swings cluster with the Support & Resistance Finder is one way to check for this kind of overlap. Overlap matters because people who read charts in different ways end up placing orders in the same price zone. A zone where price paused before is also the break-even price for those who bought then, so buying to defend it is thought likely when price comes back down. When checking, look at the zone around a level broadly rather than at its exact number, and watch how volume changes in that zone. A level with no overlapping evidence at all may draw that much less attention.

A sober look at predictive power

The evidence that Fibonacci levels have any special predictive power is weak. For a start, the levels are tightly packed. The five lines from 23.6% to 78.6% sit every 12 to 17% of the swing, so wherever a pullback stops, there is bound to be a level nearby. Even if you count only 3% of the swing above and below each of the five levels as 'near', 30% of the range from 0% to 100% is near one level or another. When many people watch the same line, orders can gather at that price and it can actually stop there, but that is closer to a result of people's habits than a property of the ratio. How often price bounced at the levels was not measured in this article. So it is more honest to use Fibonacci levels not as a prophecy that price will stop here, but as a scale that describes, with the same yardstick, how far a pullback has come. They are useful for describing a state, such as 'it has retraced about half the swing' or 'it has passed 61.8% and is heading toward 78.6%'. Rather than deciding in advance whether price will stop at a level, many people watch how the shape of the bars and volume change near it.

How it looks different in crypto and stocks

Crypto trades around the clock, so highs and lows run continuously from bar to bar, and the start and end of a swing are relatively clear to pick. Stocks build up news after the regular session closes, so the next open can start far from the previous close; no trading took place in the price zone skipped by such a gap, so a level that falls inside it has a hard time acting as a reference point. If a gap sits at the start or end of a swing, it also becomes unclear which bar's wick to use as the point. Korean stocks have a daily price limit, so the high on a day that hits the limit may be a price set by the rules rather than by demand, while US stocks have no daily limit. Volatility differs a lot as well. Crypto can multiply in price within weeks, so the same level can appear in quite different places on a linear and a log scale. Some charting programs offer an option to calculate levels on a log scale, so when comparing with someone else's levels, check which scale they are based on. Large caps such as Samsung Electronics and SK Hynix, and US indexes, move relatively gently, so the difference between the two scales tends to be small.

Reading it on a live chart

On a live chart, the end point of the swing may be a bar that is still forming. With a method like this site's tool, which picks the highest high and lowest low in the range automatically, every time the forming bar makes a new high, the 0% level rises and all the other levels move with it. A price that was at 61.8% a moment ago can be near 50% a few minutes later. And if the forming bar's wick touches a level and comes back before the close, the 'level touch' you saw on screen turns into an entirely different shape on the closed bar. On a live chart, first check whether both points of the swing are closed bars, and judge the reaction near a level after the bar has closed, so that what you see matches the record.

A practical checklist

Following the order below each time you draw levels reduces fitting after the fact. Each step is for organizing your reading and guarantees nothing about the outcome.

• Decide on the bar length and the scale (linear or log) first
• Set the rule for picking the swing's two points in advance (wick tips or bodies, and how far the range goes)
• Check that both points are on closed bars
• Make sure 0% sits at the end of the swing and write down the level values
• See whether a previous high or low, a moving average or a heavy-volume zone overlaps a level
• Judge the reaction at a level only after seeing the bar close and the change in volume