October 30, 2026 6 min read The Chartix Editorial Team

A hand on a mouse in a dim room, a monitor behind it where a glowing cyan line rises through a pale horizontal band and falls back

Price closes through a level that has held for weeks. It looks decisive. Two sessions later it is back inside the range, and everyone who read the break as a new direction is now on the wrong side of it. This happens often enough that “fakeout” has become its own word, and the pages that rank for it mostly offer tricks to avoid one.

A false breakout is a move through a support level, resistance level or pattern boundary that fails to hold, with price returning to the range it broke out of. Whether a breakout is “false” depends entirely on the rule that defined it (wick or close, how far, for how long), so the useful question is which rule you are using.

That definitional point is where most explanations go wrong, so it comes first. After it: what the measured data says about breaks that fail, why stop orders make failed breaks so common at round numbers, and the price that ends a breakout read.

A false breakout is only false relative to a rule

There is no such thing as a false breakout in the abstract. There is a breakout by some definition, followed by price doing something that definition did not expect. Change the definition and the same chart has a different number of false breakouts.

Three choices define every breakout rule:

Wick or close. Does price trading through the level count, or only a close beyond it? An intraday poke through resistance that closes back below it is a failed breakout under a wick rule and no breakout at all under a close rule. The confirmation candle post covers why the close is the standard for most pattern definitions.
Distance. Does any close beyond the level count, or must it close some distance beyond, such as a percentage of price or a multiple of the typical daily range?
Time. Does one close count, or must price stay beyond the level for several sessions?

Each filter removes some failures, and each one does it the same way: by waiting. A close rule reacts later than a wick rule. A distance filter reacts later than a bare close. A time filter reacts later still. There is no filter that removes failed breaks without also making every genuine break register later, and further from the level. That trade-off is the whole subject. Any page that promises a filter with no cost has not described what it costs.

What does the measured data say about breaks that fail?

The most specific public measurements come from Thomas Bulkowski, who records how often chart pattern breakouts fail and reverse. His definition of a busted pattern is precise: price breaks out, moves no more than 10%, reverses, and closes beyond the opposite side of the pattern.

For the symmetrical triangle, in his bull-market data, patterns with downward breakouts bust 48% of the time (his busted-pattern page does not state the sample size or period for that figure). Nearly half of the downward breaks from that shape, by his definition, failed and reversed through the top. He also notes a tendency for symmetrical triangles to double bust, with the reverse breakout failing too, though on the same page only 6% of busted triangles double bust against 67% that single bust.

His break-even failure rate is a softer measure of the same thing: the share of patterns that fail to move at least 5% in the breakout direction. For symmetrical triangles it is 25% for upward breakouts and 37% for downward ones in his bull-market figures (based on what he describes as over 3,000 trades; the page does not state the period).

These figures are one analyst’s counts on his own sample and his own definitions. They do not transfer to another market or timeframe. What they do establish is that failed breakouts are not rare accidents. On the most detailed public data, they are a routine outcome.

Why are breaks at round numbers so often short-lived?

Part of the answer is in where traders place their orders. Carol Osler, then at the Federal Reserve Bank of New York, examined almost 9,700 stop-loss and take-profit orders placed at a large foreign exchange dealing bank between September 1999 and April 2000, in dollar-yen, dollar-pound and euro-dollar. The requested prices were strongly clustered at round numbers, and the two order types clustered differently:

Take-profit orders were more strongly clustered at round numbers than stop-loss orders, and because they are executed against the direction of the move that reaches them, they tend to stall a trend at those levels.
Stop-loss buy orders clustered just above round numbers, and stop-loss sell orders just below them. When price crosses a round number, it triggers a cluster of stops that pushes it further in the same direction.

Her conclusion was that this order structure helps explain two familiar observations from technical analysis: trends tend to stall at support and resistance levels, and they tend to gain momentum once those levels are crossed. Her earlier study, published by the New York Fed in 2000, tested support and resistance levels published daily by six firms (commercial banks, investment banks and real-time information providers) from January 1996 to March 1998, and had found those levels were useful for predicting where intraday currency trends would be interrupted.

Osler’s finding is about momentum after a level is crossed, not about failure. The link to false breakouts is an inference, and worth labelling as one: if a move through a round number is driven mainly by a burst of triggered stops rather than by new participants with a view, then once that burst is exhausted there may be little left to carry price further. In that case the break was real, but the participants behind it were mostly people being forced out. The support and resistance post covers why round numbers attract levels in the first place.

This is evidence from one bank’s currency order book over seven months. It describes a mechanism, not a rate, and it should not be stretched into a claim about how often stock breakouts fail.

What makes a breakout harder to dismiss?

The tradition offers several checks. None is a guarantee, and each should be read as a way to describe the break more completely, not as a filter that sorts real from false:

The close, not the wick. A close beyond the level means price held there at the moment most participants mark their positions.
Volume relative to average. A break on volume well above the instrument’s recent average involved more participants. The volume post covers why that makes a move harder to dismiss without telling you its direction. Participation is not the same as follow-through, though: Bulkowski’s own volume resource notes a study finding that failures tripled after high-volume breakouts (the page does not state that study’s sample, market or period).
The level’s history. A level that was tested several times and held is a more meaningful thing to break than one drawn through two points.
Where the break happened in the pattern. Bulkowski reports that symmetrical triangle breakouts occur on average about 74% of the way to the apex. A break very late in a pattern, when the range has nearly vanished, is breaking out of very little.

What ends a breakout read?

Every breakout read has an invalidation, and it should be named at the moment of the break, not after:

For a break above resistance: a close back below the level that was broken, and more decisively, a close back below the low of the range or pattern it broke out of.
For a break below support: a close back above the broken level, and more decisively, a close back above the high of the range or pattern.

The first line puts the read in doubt. The second ends it. Bulkowski’s busted-pattern definition uses the far side of the pattern for exactly this reason: a return to the level is common, while a close through the opposite side is a different event.

What this post is not

It is not a method for avoiding false breakouts, because there is no method that avoids them without also delaying every real one. It is a way to describe a breakout honestly: the rule that defined it, the participation behind it, the level it broke and how often that level had held, and the close that would end the read. If you cannot state the rule before the break, any break can be relabelled false afterwards, and nothing has been learned.

For the levels themselves, start with how to read a stock chart.

Sources

Every figure above was checked against these before publishing.