Fibonacci retracements overlay horizontal levels on a price move at ratios derived from the Fibonacci sequence - most commonly 38.2%, 50%, and 61.8% of the prior swing. The idea is that pullbacks tend to stall and reverse at these levels. The tool is everywhere. Whether it has real predictive power is a fairer question than most chartists admit.
Why they sometimes "work"
The honest explanation is largely self-fulfilling: enough traders draw the same levels and place orders there that the levels become real - not because of any cosmic ratio, but because of clustered orders. A 50% retracement is just "halfway back," which is a natural place to expect a bounce regardless of Fibonacci.
The skeptic's case
You can fit Fibonacci levels to almost any move after the fact, and the "hits" get remembered while the misses get forgotten. Drawn on different swings, they give different levels. That flexibility is the tell of a tool that describes the past better than it predicts the future - the opposite of what a systematic edge requires.
Fib levels work best as a place other traders are watching - not as magic numbers.
A grounded use
If you use them at all, treat Fib levels as one more spot where orders may cluster - useful only when they line up with real support and resistance, prior-day levels, or VWAP. When a Fib level and a structural level coincide, the confluence matters. The Fib line by itself is a suggestion, not a signal. Data and structure beat ratios drawn from a rabbit-breeding sequence.