Fakeout

Price pushes through a level that a chart reader was already watching, a support or resistance line, the boundary of a range or the edge of a pattern, and then returns to the side it came from. The name is applied after the return, never before it: while the bar is forming, a fakeout and a genuine breakout are the same object on the same chart, and the only thing that separates them is what happens next.

There is no standard test for the return. Some desks require a close back inside the level on the timeframe the level was drawn on, others allow a fixed number of bars, others wait for price to reach the far side of the prior range before calling it. Those conventions disagree with one another, and on any given chart two of them can be satisfied and a third not, so the same sequence is honestly describable as a fakeout by one reading and as a shallow pullback by another.

The mechanism most often cited is that resting orders cluster just beyond an obvious level, so a modest amount of buying or selling can carry price through it, trigger those orders, and then find nothing behind them to continue on. That makes a fakeout a statement about where orders were sitting rather than about what an instrument is worth. The thing that most often trips people up is the hindsight problem: a record of fakeouts collected after the fact contains only the breaks that failed, which says nothing about how often a break that looked identical carried on.