Fibonacci retracements: how to draw them, what the levels mean and where they fall short

A Fibonacci retracement is a drawing tool that takes a completed price move, from a swing low to a swing high or the reverse, and marks set percentages of that move (usually 23.6%, 38.2%, 50%, 61.8% and 78.6%) as possible areas where a pullback could pause or turn. Traders use the levels to look for entries in the direction of the original move and to decide where a trade idea is wrong. The levels are not magic: they are a consistent way to divide a swing, and they work only as well as the rest of your analysis.

Where the numbers come from

The Fibonacci sequence is named after Leonardo of Pisa, known as Fibonacci, who introduced it to European readers in his book Liber Abaci in 1202 (the sequence was known earlier in Indian mathematics). Each number is the sum of the two before it: 0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144, and so on.

The ratios traders use come from dividing numbers in the sequence:

• Divide any number by the next one, and as the numbers get larger the answer settles at about 0.618 (for example, 89 ÷ 144 = 0.618). That is the 61.8% level.
• Divide a number by the one two places further on, and you get about 0.382 (55 ÷ 144 = 0.382). That is the 38.2% level, and it also equals 1 minus 0.618.
• Divide a number by the one three places further on, and you get about 0.236 (34 ÷ 144 = 0.236). That is the 23.6% level.
• The 78.6% level is the square root of 0.618. It is a later addition that some platforms show by default and some do not.
• The 50% level is not a Fibonacci ratio at all. It is included because traders have long watched halfway retracements, an idea associated with Dow theory, and because it sits neatly between 38.2% and 61.8%.

The inverse of 0.618 is about 1.618, the so-called golden ratio, which also appears in extension levels such as 161.8%. You will read a lot about the golden ratio appearing in shells, flowers and the Parthenon. Some of those claims are true, many are exaggerated, and none of them is evidence about currency prices. A ratio that describes the spiral of a sunflower head tells you nothing about how banks and funds will trade EUR/USD next Tuesday.

The use of these ratios in markets owes a lot to Elliott wave theory, developed by Ralph Nelson Elliott in the 1930s, which leaned heavily on Fibonacci relationships. Today the tool is built into MetaTrader, cTrader, TradingView and most other charting packages, which is a large part of why it is so widely watched. Our guide to cTrader and TradingView covers those platforms.

How to draw a retracement

The mechanics are simple. The judgement is not.

1. Find a clear, completed swing. In an uptrend, that is a move from a swing low up to a swing high that has since started to pull back. In a downtrend, it is a move from a swing high down to a swing low.
2. Anchor the tool. In an uptrend, click the swing low first and drag to the swing high. In a downtrend, click the swing high first and drag to the swing low. Your platform then draws horizontal lines at the chosen percentages between the two points.
3. Read the levels as pullback depth. In an uptrend, the 38.2% line is where price would sit if it had given back 38.2% of the rise. The 61.8% line marks a deeper pullback.

If the order of clicks confuses you, check that the 0% line sits at the end of the move (the high in an uptrend) and the 100% line at its start. If it is the other way round, you have drawn it backwards. Platforms differ slightly in how they label the lines, so look before you rely on them.

Which swing?

This is where most of the subjectivity hides. A chart contains swings within swings, and the "right" one depends on the timeframe you trade and how long you intend to hold. A few practical guidelines:

• Choose a swing that most people looking at the same chart would agree is obvious. If you need to squint to find it, other traders probably will not be watching it.
• Use the timeframe you actually trade, then check the next timeframe up for a larger swing that might matter more. Our guide to multiple timeframe analysis explains how to combine the two.
• Decide your anchoring rule in advance and stick to it. Changing the anchor points until the levels line up with something on the chart is the commonest way people fool themselves with this tool.

Wicks or bodies?

Most traders anchor at the extreme wick high and low, because that is the actual highest and lowest traded price on that chart. Some anchor at candle closes, arguing that closes reflect where the market settled. Either is defensible. Remember too that most retail charts are drawn from bid prices and that different brokers' feeds differ slightly, so the extremes on your chart may not match someone else's to the pip. If the bid/ask distinction is unfamiliar, see reading a forex quote.

Worked example: drawing the levels

This is an illustrative example with made-up prices, to show the arithmetic, not a view on any market.

Suppose EUR/USD rose from a swing low at 1.0600 to a swing high at 1.1000 and has started to pull back. The swing is 400 pips (if pips are new to you, start with what is a pip). Each retracement level is the high minus that percentage of the swing:

• 23.6%: 400 × 0.236 = 94.4 pips below the high, about 1.0906
• 38.2%: 400 × 0.382 = 152.8 pips, about 1.0847
• 50%: 200 pips, 1.0800
• 61.8%: 400 × 0.618 = 247.2 pips, about 1.0753
• 78.6%: 400 × 0.786 = 314.4 pips, about 1.0686

Now change one decision. Suppose the low candle had a wick down to 1.0580, and you choose to anchor at the wick rather than at the 1.0600 area where most candles bottomed. The swing is now 420 pips and the 61.8% level moves to about 1.0740, 13 pips lower. Anchor the high at a 1.1020 wick as well, and it moves again, to about 1.0748.

Thirteen pips may not sound like much, but on a trade with a 40-pip stop it is a third of the risk. And this is the simple case: if you had chosen a different swing altogether, the levels could sit 50 or 100 pips away from these. This is why it makes more sense to think of Fibonacci levels as zones than as lines, just as with support and resistance.

Extensions: levels beyond the swing

The same tool can project levels past the end of the swing. In the example above, the 127.2% and 161.8% extension levels would sit at 1.0600 plus 1.272 and 1.618 times the 400-pip swing: about 1.1109 and 1.1247. Some traders use these as areas to take some or all of a position off if price makes a new high.

Be clear about what this is: a way of choosing exit areas in advance, using a consistent rule. It is not a forecast that price will reach 1.1247, and there is no reason to expect a move to stop there rather than anywhere else. Different platforms also offer different extension tools (some project from the end of the pullback rather than from the swing), so make sure you know which one you are using.

Why traders think the levels work

There are two broad explanations, and they are worth separating.

The self-fulfilling argument. Because the tool is on every platform and taught in every introductory course, many traders watch the same levels on obvious swings. If enough of them place orders near 61.8% of a big daily move, those orders can cause a reaction there. This is real in the same limited way it is real for round numbers and old highs. It is strongest on the most obvious swings on higher timeframes, and weakest on swings that only you can see.

The pullback argument. Markets that trend tend to move in steps: a push, a pullback, another push. Pullbacks have to end somewhere, and most of them end somewhere between a quarter and three-quarters of the previous move. That tells you something about how trends behave, but it does not say anything special about 61.8% compared with 60% or 65%.

The limits, stated plainly

We are not aware of convincing evidence that markets respect Fibonacci ratios more than other percentages, and you should treat confident claims either way with suspicion. What can be said with certainty is that several features of the tool make it look better than it is.

There is nearly always a level nearby

The standard levels at 23.6%, 38.2%, 50%, 61.8% and 78.6% are close together. The largest gaps are 11.8 percentage points (between 38.2% and 50%, and between 50% and 61.8%) and 16.8 points (between 61.8% and 78.6%). So a pullback that stops anywhere between 23.6% and 78.6% is never more than about 8.4% of the swing away from a Fibonacci level. On the 400-pip swing above, that is at most about 34 pips, and usually much less. Add a zone around each level and almost every turning point will appear to "respect" one of them.

Hindsight makes it look precise

After the event, it is easy to find the swing that makes the turning point line up with a level. In real time, before the pullback ends, you do not know which level will be the one, or whether price will go straight through all of them and make a new low. A chart that shows a perfect bounce at 61.8% does not show the four other levels price might have bounced from, or the many charts where it did not bounce at all.

No pattern works reliably on its own

A Fibonacci level is a price derived from two other prices. It knows nothing about interest rates, positioning, liquidity or the data due tomorrow. A surprise from a central bank or an inflation release can move a currency through every level on the chart in minutes; see how economic data moves forex. Thin markets and weekend gaps can do the same, as gap risk and weekend risk explains.

Confluence: useful, but count honestly

Most traders who use Fibonacci levels look for confluence: a retracement level that coincides with something else, such as a previous support or resistance zone, a round number, a moving average or the prior week's low. The logic is sound as far as it goes: if several independent reasons point to the same area, more traders may be watching it.

The catch is the word independent. If you have five Fibonacci levels, a dozen horizontal levels, three moving averages and every round number on the chart, almost any price will line up with two or three of them. Confluence only means something if you defined the ingredients before looking for the overlap, and if the list is short. A sensible rule of thumb is to require at most one or two other factors, and to be able to name them before price arrives.

Moving averages as a second ingredient are covered in trend following and moving averages. Some traders also look at momentum readings near a level; RSI and MACD explained covers why those add less independent information than they appear to.

Worked example: planning risk around a retracement

Again, illustrative numbers only, continuing the EUR/USD swing from 1.0600 to 1.1000. Suppose the 61.8% level at about 1.0753 also sits close to an old resistance zone around 1.0740 to 1.0760 that price broke on the way up. A trader looking for a long position might plan:

• Entry: 1.0755, inside the area where the 61.8% level and the old zone overlap.
• Stop-loss: 1.0670, below the 78.6% level at 1.0686 with a buffer of about 16 pips for noise and spread.
• Risk: 85 pips.

Now the sizing. The account is EUR 10,000 and the trader risks 1% per trade, so EUR 100. On EUR/USD a mini lot (10,000 units) is worth about $1 per pip, so 85 pips is about $85 per mini lot, or roughly EUR 79 at the illustrative exchange rate. That allows about 1.2 mini lots (0.12 lots), putting roughly EUR 95 at risk. The method is set out in position sizing, and the position size calculator does the arithmetic for you.

If price returned to the swing high at 1.1000, that would be 245 pips from entry, almost three times the risk. That is a description of the possible outcome if things went well, not a target. Price may stall at the 38.2% level on the way, or never get back above the entry at all. How often trades like this reach their exit is something only your records can tell you; see risk-reward ratio for the maths of win rates and expectancy.

A second trader might reason differently: if the uptrend is real, price should not go back below the swing low at 1.0600, so the stop goes at 1.0585. That is a 170-pip risk, so the same EUR 100 allows only about 0.06 lots. Neither approach is right or wrong. A tighter stop gives a larger position and more frequent stop-outs; a wider stop gives a smaller position and fewer. The where to put a stop-loss guide works through that trade-off in detail.

Notice what the Fibonacci tool contributed: a consistent way to choose an entry area and a stop area. It did not contribute any knowledge about whether price would turn.

How to test whether the levels help you

If you want to know whether Fibonacci levels add anything to your trading, the honest test is a comparison. Write a rule, for example "in a daily uptrend, enter long at the first touch of the 61.8% level of the latest swing, stop beyond 78.6%, exit at the swing high". Then run the same rule with arbitrary percentages, such as 60% and 75%, or 55% and 70%. If the Fibonacci version does not do clearly better across a large sample, after costs, the ratios are not doing the work you thought.

You can do this manually or with code; how to backtest a strategy covers both, along with the overfitting trap of trying many anchoring rules until one looks good. Then keep a trading journal of live or demo trades that used the levels, including the ones where price went straight through them. Most traders only remember the bounces.

Common mistakes

• Redrawing until it fits. Moving the anchor points after the fact to make a turning point line up with a level teaches you nothing and builds false confidence.
• Treating levels as exact prices. Price overshoots by a few pips all the time, and your broker's feed may differ from the chart you learnt on. Use zones.
• Drawing on tiny swings. A retracement of a 20-pip move on a one-minute chart is mostly noise, and the spread is a large fraction of the distance between levels. The costs side is explained in spreads and commissions.
• Using every level as a reason to trade. If 38.2% is a buy, and 50% is a buy, and 61.8% is a buy, the tool is no longer filtering anything.
• Ignoring the calendar. A neat retracement into a level an hour before a central-bank decision is a different trade from the same picture on a quiet day. Check how to use an economic calendar.
• Sizing up because the confluence looks strong. Confidence in a chart does not change what leverage does to losses; see leverage and margin.
• Forgetting that the trend can end. A pullback that passes 78.6% may simply be the start of a reversal. Have an exit for that case before you enter.

Keep the risk in proportion

Fibonacci retracements are a tidy way to divide a swing and plan around it. They do not change the odds of trading in your favour. Most retail CFD accounts lose money (brokers' required disclosures commonly show figures between about 61% and 89%), and leverage magnifies losses. Our guide to why most retail CFD traders lose explains what those disclosures show. For a short definition you can share, see the glossary entry for Fibonacci retracement.

Frequently asked questions

What are the main Fibonacci retracement levels?

The common levels are 23.6%, 38.2%, 50%, 61.8% and 78.6% of a completed swing. The 23.6%, 38.2% and 61.8% levels come from ratios between numbers in the Fibonacci sequence, 78.6% is the square root of 0.618, and 50% is not a Fibonacci ratio but is included because traders have long watched halfway pullbacks.

How do you draw a Fibonacci retracement correctly?

In an uptrend, anchor the tool at the swing low and drag to the swing high; in a downtrend, from the swing high to the swing low. Choose an obvious, completed swing on the timeframe you trade, decide in advance whether you use wicks or closes, and do not move the anchors afterwards to make the levels fit.

Which Fibonacci level is the most important?

Many traders pay most attention to 61.8% and 50%, and some to 38.2% in strong trends. There is no level that price reliably respects, and claims that one is "the" level usually come from looking at charts after the fact.

Do Fibonacci retracements work in forex?

They work as a consistent way to divide a swing and plan entries and stops, and they may attract some reaction on obvious swings because many traders watch them. They do not predict where a pullback will end, and because the levels are close together, price is near one most of the time anyway. Test any rule against arbitrary percentages before trusting it.

What is the difference between Fibonacci retracements and extensions?

Retracements measure how far price has pulled back within a completed swing. Extensions project levels beyond the end of the swing, such as 127.2% and 161.8%, and some traders use them to plan exits. Neither tells you where price will go.

Should I use Fibonacci levels on their own?

No. A level derived from two prices has no information about news, liquidity or positioning. Treat it as one input, combined with market structure and a clear plan for where you are wrong, and size every trade so that being wrong costs a small, known amount.