Enter a swing high and a swing low and you get the retracement and extension tables together. Every ratio shown is sourced, and the 50% level carries a footnote saying it is not a Fibonacci number at all.
Where the Fibonacci ratios come from
Three of the ratios shown derive directly from the Fibonacci sequence by dividing its terms. The 61.8% ratio is a term divided by the next one, 38.2% is a term divided by the one two places further along, and 23.6% is a term divided by the one three places further along. The 78.6% level is not a division of terms at all but the square root of 0.618, which Corporate Finance Institute states explicitly. The ratios are therefore not an arbitrary choice but the outputs of one repeated arithmetic operation.
The 50% level and the exception nobody mentions
The 50% level is not from the Fibonacci sequence at all, and this is not a footnote-sized point. StockCharts states that this number is not based on a Fibonacci number and instead stems from Dow Theory assertion that averages often retrace half their prior move. CME Group says the same: the third common ratio is 50%, which is not a number from the Fibonacci series but is included in the analysis. That is why the 50% row in the table above carries a permanent footnote.
Retracement: how and from where it is measured
Retracement is measured against the swing range, the gap between the high and the low. In an uptrend the percentage is subtracted from the high, because the move rose and the retracement falls. In a downtrend it is added to the low, because the move fell and the retracement rises. The convention used here is printed under the table: 0% marks the end of the measured move and 100% marks its start. Sources differ on naming rather than on prices, so we printed the convention instead of picking a winner.
Worked example: an uptrend on EUR/USD
Take EUR/USD rising from a low of 1.08100 to a high of 1.09200. The swing range is 0.01100. The 23.6% level sits at 1.08940, 38.2% at 1.08780, 50% at 1.08650, 61.8% at 1.08520 and 78.6% at 1.08335. Each is the high minus the range multiplied by the ratio. The band between 38.2% and 61.8% is shaded in the table because it is the most watched, not as a label implying quality or a better signal.
Worked example: the same data in a downtrend
Now the same data in a downtrend. The 23.6% level moves to 1.08360, 38.2% to 1.08520, 61.8% to 1.08780 and 78.6% to 1.08965. Note something worth pausing on: the 50% level stayed at 1.08650 in both directions. The reason is that it is exactly the midpoint of the range, and a midpoint is the same whether you measure from the top or the bottom. That makes a useful sanity check: if your 50% level differs between the two directions, something was entered wrong.
Extensions and their two constructions
Extensions are built in two published ways. The two-point construction projects the swing itself beyond its end at 127.2%, 161.8%, 200%, 261.8% and 423.6%. The three-point construction starts from the end of the retracement and adds the move multiplied by the ratio. Naming differs between platforms: what TradingView calls a trend-based Fib extension is what MetaTrader and cTrader call an expansion, and they are the same construction.
Worked example: the gap between the two modes
The gap between the two modes is not theoretical. On the same data — a low of 1.08100, a high of 1.09200 and a retracement ending at 1.08650 — the two-point construction puts the 161.8% extension at 1.09880, while the three-point construction puts it at 1.10430. The gap is 0.00550 and it is constant across every ratio, because the two modes start from anchors that differ by exactly that amount. That is why the mode is printed in the extension table heading rather than left implicit.
The discretionary part: picking the high and low
The most delicate part of this tool is not the arithmetic but what precedes it. The calculator is exact on whichever two points you give it, yet choosing the high and the low is itself a judgement call: which timeframe you measure on, which swing you treat as the main move, and whether you use the wick extreme or the candle body. Two traders looking at the same chart can pick different points and get completely different levels, and both used the calculator correctly.
Common mistakes
Four mistakes recur. The first is mixing the two modes without noticing, so a 161.8% extension computed one way gets compared with one computed the other. The second is using unverified ratios such as 138.2%, which appears in no acceptable source and which we therefore excluded. The third is inverting the high and the low, an error the calculator blocks with an explicit message. The fourth is assuming a level will necessarily halt price; a level is a computed number, not a barrier.
Frequently asked questions
Is the 50% level a Fibonacci ratio?
No. Three reference sources state this directly: the StockCharts reference writes that 50% is not based on a Fibonacci number and stems from Dow Theory, CME’s education page writes that it is not from the Fibonacci series but is included in market analysis, and CFI calls it a common convention. It appears here because traders use it widely, not because of a mathematical origin, and we say so beside the row itself rather than in a distant footnote.
Why does a 161.8% extension give me two different prices?
Because there are two published constructions with different anchors. The two-point construction builds the extension from the swing high and low alone; the three-point construction builds it from the end of the retracement plus a percentage of the original move. TradingView calls the second one a ‘Trend-Based Fib Extension’ and requires three clicks, while MetaTrader and cTrader call the same construction an ‘Expansion’. Same percentage, different anchor, different price.
Why do you not show the 138.2% level?
Because we could not find it in any acceptable reference source. We checked TradingView, MetaTrader 5, cTrader and Sierra Chart documentation, the StockCharts reference, the BabyPips school, and the Schwab, Fidelity, IG, CME and CFI pages. The level appears only on tool sites and blogs, and our rule is that a formula source is never a tool site. The closest support we found is CME listing the value 1.38 inside a set it describes as ‘the same as the retracement ratios’, which is not enough to publish it as a standard level.
Which end is 0% and which is 100%?
Sources disagree on the labels, not on the prices. The BabyPips reference puts 100% at the low in an uptrend, while cTrader’s documentation puts 0% at the start of the move, which is the high in a decline. The resulting prices are identical either way; only the endpoint labels flip. We use 0% at the end of the measured move and 100% at its start, and we print that convention right under the table heading so it is never left implicit.
How do I pick the right swing high and low?
That is the discretionary part of the whole tool. The calculator computes precisely on the two points you give it, but it does not choose them for you. Traders normally pick a clear high and low on the timeframe they work with, and a different choice produces entirely different levels from the same chart. This is why two traders looking at the same chart can end up with different Fibonacci levels.
Do these levels tell me when to buy or sell?
No. The calculator derives prices from two or three points you entered; it issues no entry or exit recommendation and forecasts no direction. The levels mark zones many traders watch, and nothing guarantees price will react at them. The decision and the risk management are yours, and position size comes from how much you accept losing, not from where a level sits.