What counts as a chart pattern failure?

Definitions vary, so any quoted rate depends on the rule behind it. Bulkowski's benchmark counts a failure when price breaks out and reverses before achieving a five percent move; other researchers use a close back inside the pattern, a stop-distance loss, or a missed measured target. A published failure rate is only meaningful alongside the definition that produced it.

Can I rely on published pattern failure rates?

Treat them as rough priors, not forecasts. Base rates were measured on specific markets, timeframes, and eras, and they drift with volatility regime and sample choice. They are most useful comparatively (which patterns and directions have historically failed more often) and for forcing you to define failure before entry. They do not make any single pattern's outcome predictable.

Which chart patterns fail most often?

The honest answer is that rankings depend on the study's definitions, but the structural findings repeat: countertrend breakouts fail more than with-trend ones, late apex-area resolutions underperform earlier exits, and patterns traded against the larger regime degrade everywhere. Quoting a single percentage without its methodology transfers false precision; the comparative rankings travel better than the numbers.

Do Bulkowski-style statistics still hold today?

Partially, and that is the expected outcome. His samples skew toward late-twentieth-century US equities, and markets since have changed structure, tick sizes, and participant mix; some documented tendencies persist in newer samples, others faded. The durable inheritance is the method, define, count, compare, which any trader can re-run on current data rather than debating old tables.

Why do average pattern moves fall short of measure-rule targets?

Because the measure rule is a geometric convention, not a measured expectancy. Counting studies consistently find the mean post-breakout move smaller than the projected height, with wide dispersion around it. The practical adjustment is standard: treat the textbook target as a ceiling scenario, plan around the documented average, and let management rules capture the outliers.

How large a sample do pattern statistics need?

Hundreds of instances per pattern-and-context cell before rates stabilize, which is why serious counts span many symbols and years. Each conditioning split (direction, trend, regime) divides the sample again, and rare patterns may simply never accumulate enough cases for their statistics to mean much, itself a finding worth respecting when someone quotes precise rates for an exotic formation.

Concept

Pattern Failure Statistics

Pattern Failure Statistics, also known as fakeout frequency, apex behavior, maturity/timeout, are Chart & Candlestick Patterns concepts.

Bulkowski base rates

What are Pattern Failure Statistics?

Pattern failure statistics are measured base rates for how often a chart pattern actually does what the textbook says. Instead of treating a formation as a picture, the approach defines an outcome in advance (does the breakout follow through, how far does price travel versus the measure rule target, how often does it reverse immediately) and counts results across a large historical sample. Thomas Bulkowski's catalog work popularized the method, including the widely used benchmark that counts a breakout as failed when price reverses before moving even five percent in the breakout direction.

The measuring tradition is younger than the patterns. Classical authors described formations qualitatively; systematic counting arrived with computing power, from Bulkowski's encyclopedic samples published from 2000 onward to academic work such as the Lo, Mamaysky and Wang study that formalized pattern detection with statistical filters and tested whether the shapes carried any information at all. The shared conclusion across serious efforts: patterns are conditions with measurable, modest, regime-dependent tendencies, not promises.

The findings that recur across studies are structural rather than numeric: failure rates differ by pattern, by breakout direction, and by market regime; average moves usually fall short of textbook targets; and timing matters, for example triangle breakouts that arrive very late, near the apex, tend to be weaker than ones that leave earlier. A pattern that exceeds its typical duration without triggering is treated as stale. Published numbers shift with the sample and the era, so they are priors, not promises.

The methodology is where quoted numbers earn or lose credibility. A failure rate depends on the pattern's detection rules (hand-labeled shapes differ from algorithmic matches), the outcome window, the failure threshold, and the sample's era and universe; survivorship-biased stock samples flatter bullish patterns, and one volatility regime's rates mislead in another. Building statistics for your own market, with rules defined before counting, is worth more than any borrowed table, and it is the only way the numbers reflect the patterns as you actually trade them.

How to measure pattern failure statistics

The exercise is measurement design; every step exists to keep the count honest.

1Define the pattern objectively: rules a scanner could apply (touch counts, boundary slopes, body ratios), because hand-picked examples smuggle in hindsight.
2Define the outcome and the failure in advance: the breakout direction, the follow-through threshold, the target convention, and the time window.
3Sample widely: many symbols, many years, spanning more than one volatility regime, with delisted names included where possible.
4Record distributions, not just averages: the spread of outcomes and the frequency of immediate reversals matter more than a mean move.
5Split by context: direction, prevailing trend, and timing within the pattern (early versus apex exits in triangles) routinely change the rates.
6Hold out data: rates tuned and verified on the same sample are decoration; the out-of-sample count is the statistic.

How traders use it

To predefine failure: a false breakout threshold (a reversal before some minimal move, or a close back inside the pattern) is set before entry, so stop placement follows from the statistic rather than from hope.
To temper targets: measure-rule projections are discounted toward the documented average move instead of being taken at face value.
To filter setups: contexts with better documented base rates, such as breakouts aligned with the prevailing trend or backed by breakout confirmation, are preferred over raw pattern matches.
To size positions: base rates translate into expectancy arithmetic, so a pattern with a documented one-in-three immediate-failure rate gets sized for that frequency rather than for the textbook picture.
To grade scanners: automated detectors for triangles, double tops and bottoms, or candlestick events are audited by their measured outcomes, which quickly separates definitions that find tradable structure from definitions that find shapes.

Failure statistics vs the pattern catalogue

Candlestick Patterns: The catalogue names shapes and tells their stories; failure statistics attach numbers to the stories. The two are complements: a pattern without a base rate is folklore, and a base rate without a defined pattern is a number about nothing.

Double Top/bottom: A staple example of why measurement matters: studies repeatedly found that strict double tops confirm far less often than casual chart-reading assumes, and that outcomes hinge on the confirmation rule chosen. The pattern's literature is now inseparable from its statistics.

Ascending/descending/symmetrical Triangle: Triangles supplied one of the tradition's structural findings: resolution timing changes outcomes, with late, near-apex breaks measurably weaker. That is a statistic about behavior inside the pattern, something no static picture communicates.

Concept family

Chart & Candlestick Patterns

84 concepts mapped · 84 in the Library