Fibonacci Retracement on EURUSD: What 95 Real Pullbacks Actually Did

Fibonacci levels on EURUSD are a map of where price is likely to pause, not a map of where it is likely to turn. On three months of live data, the deeper the retracement, the worse the odds of continuation. If you trade the 61.8% bounce with no other confirmation, you are taking a setup that failed three times out of four.

Every fib article shows a chart where price kissed 61.8% and flew. That chart exists. So do the ones nobody screenshots. I decided to count all of them instead of picking one.

How the test was run

These are real EURUSD 5-minute bars, 16,926 of them, covering June 3 to August 26, 2026.

• Impulse detection: swing highs and lows confirmed by 12 bars on each side, alternating high to low, so every leg is a genuine directional move.
• Minimum size: 25 pips. Anything smaller is noise on this timeframe.
• Measurement window: 15 hours after the leg completes, or until price retraced 100%, whichever came first.
• Continuation: price making a new extreme beyond the end of the impulse after the retracement touched the level.

That gave 95 impulses to measure. Small enough to state honestly, large enough to kill a myth.

Where pullbacks actually stopped

The uncomfortable column

Read the right-hand column top to bottom. It goes 31.2, 29.5, 31.2, 24.0, 22.2, 22.6. It drifts down.

The standard teaching says a deep retracement into the golden zone is the high-probability entry. On this sample the opposite held: the deeper price came back, the less likely the original move was to resume. Which makes sense once you say it out loud. A pullback that eats 78% of the impulse is not a pullback, it is the market changing its mind.

How I use the levels now

• 50% is the working level, not 61.8%. It gets touched two times out of three and continuation odds are as good as anywhere on the grid.
• Never enter on the touch alone. A level touch is a 24% to 31% event. It needs a trigger: a closed rejection candle, a structure break on the lower timeframe, or a momentum tool that agrees.
• Place the stop past 100%, not past 78.6%. A third of moves retrace fully before deciding, so a stop just under the last fib is exactly where the noise lives.
• Expect to miss half your setups. If you only work 61.8%, roughly 47% of impulses will never come to you. That is not bad luck, it is the base rate.

Frequently asked questions

Does this mean Fibonacci does not work?

It means the levels describe where price pauses, and pausing is not turning. Used as a map of likely pullback depth they are useful. Used as a standalone buy signal they are a coin flip weighted against you.

Which level should I draw from?

From the swing that started the move to the swing that ended it, on the same timeframe you plan to trade. In this test the legs were confirmed by 12 bars on each side, which is a strict definition and removes most of the arbitrary drawing.

Is 95 impulses enough to conclude anything?

It is enough to say the 61.8% touch is not a high-probability event on EURUSD M5 this summer. It is not enough to generalise to every pair and every regime, and I am not going to pretend otherwise.

What about the 78.6% level?

Reached by 37.9% of impulses, continuation 22.2%. It is the weakest row in the table.

Do these levels work better on higher timeframes?

Possibly, and that is the next test rather than a claim. Same code, daily bars, longer window.

Method and data. EURUSD 5-minute OHLC, 16,926 bars, June 3 to August 26, 2026, retrieved from Yahoo Finance on August 26, 2026. Swings confirmed by 12 bars either side, impulses of 25 pips or more, retracement measured for 15 hours after the leg or until a 100% retrace. Continuation defined as a new extreme beyond the impulse end. 95 impulses qualified. The script is kept with the article so every figure can be re-run.