How Do You Use Fibonacci Retracement? First Fix the Endpoints, Then Check the Invalidation Conditions

Author: Thomas | Technical Analysis | October 9, 2026

Drawing Fibonacci retracement lines on a gold chart does not mean a reliable buy point has already been found. It is first of all a method for measuring the retracement ratio of a selected swing. What really needs to be explained is: why these two endpoints were chosen, what happened after price reached the zone, and under what circumstances the original judgment becomes invalid.

Define the swing first, then discuss the ratio

When two analysts use the same tool but get different positions, it is often not a calculation error, but rather that they chose different highs and lows, timeframes, or wicks. Before analysis, first record the instrument, quote source, chart timeframe, and endpoint times, and state whether the endpoints use wick extremes or closing prices.

A pullback on a small timeframe may be only part of a swing on a larger timeframe. You cannot see which line "just happens to touch price" and then work backward to change the starting point. The selection rules need to be determined in advance, and then checked for consistency across different samples.

Educational illustration: how retracement ratios convert into prices

The following numbers are entirely an educational illustration, not today's XAUUSD market, and not a live trading record. Suppose a price segment rises from 100 to 120, with a swing range of 20; retracing downward from the high, the price position can be calculated as "120 minus 20 times the retracement ratio."

A 38.2% retracement corresponds to 112.36, a 50% retracement corresponds to 110, and a 61.8% retracement corresponds to 107.64. The three horizontal levels are only different divisions of the same range, and none of them comes with any guarantee of execution or rebound. 50% is also a common retracement level, but it should not be claimed to be the ratio of adjacent Fibonacci numbers.

Software label direction must be checked against the calculation

Different drawing directions or parameter settings may cause 0 and 100 to appear at different endpoints. Do not just look at the text "61.8" and assume it necessarily means a 61.8% retracement from the high. First compare the actual endpoints with the calculation results to confirm exactly which distance is being discussed.

MetaQuotes' official drawing tool documentation includes these retracement levels; that is an explanation of how the tool draws, not evidence that it has a fixed win rate. Being able to automatically generate horizontal lines does not mean the market has validated these levels as effective.

Only after price reaches the zone do you begin checking the reaction

If price approaches the educational level of 112.36, observe whether recognizable support subsequently forms, whether it recovers the position it previously lost, and whether this change can persist. Touching, temporarily rebounding, and forming a new structure are three different states, and a single wick cannot replace subsequent evidence.

In analysis, retracement levels can be compared with the existing price structure, but overlap is not independent probabilistic proof. Multiple tools may all use the same set of prices, and an increase in number does not automatically equal an increase in evidence. Explaining what each provides is more useful than filling the chart.

Invalidation conditions must be written clearly in advance

One research rule could be: first fix the observation timeframe and record the zone reaction; if price continues through the zone, retests fail to recover it, and lows continue to move lower, then the original support assumption needs to be reassessed. This is only a testable method example, not a universal trading parameter.

You cannot endlessly change your story every time price crosses a line and say the next one is the real support. If price returns below the starting point of the selected upward swing, the retracement explanation originally designed around that advance especially needs to be reviewed. Stop-loss, position sizing, and actual execution still must be considered separately; measurement lines cannot replace risk management.

The most misleading thing is moving the endpoints after the fact

If, during review, you first see the rebound result and then move the starting point to the position that fits best, the chart will look very accurate, but that does not mean the same judgment could have been made at the time. Keeping the first annotation, later modifications, and the reasons for modification is the only way to distinguish reasonable updates from after-the-fact selection.

When comparing methods, you must also keep samples with no reaction, rather than showing only three successful charts. Without complete records and cost statistics, do not publish so-called "golden ratio high win rate," and even less should you fabricate personal live trading returns to prove the tool works.

Put the tool back into the research process

A reviewable analysis should include endpoint rules, ratio calculations, observation zones, confirmation methods, invalidation conditions, and subsequent results. The day's levels should be separately verified using real quotes with timestamps, and the 100 to 120 illustration in this article cannot be moved into actual orders.

Fibonacci retracement is suitable for helping describe distance and organize observation; it cannot predict inevitable outcomes. Leveraged trading may cause major losses. This article is for technical research and education, does not constitute a buy or sell instruction, and does not promise returns.