fib retracements work because everyone draws them, so how much is real?
Posted: Fri Oct 30, 2026 11:14 am
Fibonacci pullbacks get sold on one claim: price retraces to 38.2, 50 or 61.8 and bounces. The less comfortable observation is how much of the pullback range those levels cover.
Draw 38.2, 50, 61.8 and 78.6 on a swing and almost every normal pullback ends near one of them. A shallow one lands at 38. A medium one at 50. A deep one at 61.8 or 78.6. Nearly any outcome counts as a hit, and the misses get explained away.
The second problem is swing selection. The start and end points are a judgment call, so two traders draw different grids on the same chart, and the grid that happened to line up is the one remembered.
Fair counterpoint: lots of traders watch these levels, so some orders cluster there, which makes the idea partly self-fulfilling. That could be enough to matter for a few pips. It just doesn't make the numbers magic.
The test is easy in principle. Record where pullbacks actually ended as a percentage of the swing, then compare that with what randomly placed levels would have scored. The result may be less impressive than the memory of the winners.
Does anyone use a fixed pullback depth rule on the 1-minute chart, and did it beat a simple "wait for structure to hold" rule when logged?
Draw 38.2, 50, 61.8 and 78.6 on a swing and almost every normal pullback ends near one of them. A shallow one lands at 38. A medium one at 50. A deep one at 61.8 or 78.6. Nearly any outcome counts as a hit, and the misses get explained away.
The second problem is swing selection. The start and end points are a judgment call, so two traders draw different grids on the same chart, and the grid that happened to line up is the one remembered.
Fair counterpoint: lots of traders watch these levels, so some orders cluster there, which makes the idea partly self-fulfilling. That could be enough to matter for a few pips. It just doesn't make the numbers magic.
The test is easy in principle. Record where pullbacks actually ended as a percentage of the swing, then compare that with what randomly placed levels would have scored. The result may be less impressive than the memory of the winners.
Does anyone use a fixed pullback depth rule on the 1-minute chart, and did it beat a simple "wait for structure to hold" rule when logged?