Chart patterns: reversals, continuations, and invalidation

A chart pattern is a way to describe a sequence of price swings. The useful part is a reproducible setup and failure condition. Recognizing a familiar silhouette is only the starting point.

Reversal candidates

Head and shoulders describes three peaks with the middle one highest; the intervening lows define a neckline. An inverse formation reverses that geometry. A double top or bottom describes two tests of a similar area separated by a reaction. Until a defined boundary breaks, the formation may remain an unfinished range.

Avoid treating every second test as a reversal entry. Define the allowed distance between peaks, minimum separation, and confirmation event. A visually convincing shape can contain too little information to justify its stop distance.

Continuation candidates

Flags describe a consolidation after an impulse. Triangles describe converging boundaries, while wedges typically have both boundaries sloping in the same direction. A cup-and-handle candidate combines a rounded recovery with a smaller consolidation. Pattern names alone do not establish the direction of the eventual breakout.

If a system labels a formation bullish, distinguish that expectation from an observed break. A descending boundary can persist much longer than an early entry can tolerate.

Measured moves are planning references

Suppose a hypothetical range spans 90 to 100. Projecting its 10-point height above a confirmed break at 100 gives a reference target of 110. That arithmetic says nothing about the probability of reaching 110. Resistance at 104, wider spreads, or a distant invalidation may make the available trade unattractive.

Keep the measured target separate from a performance forecast. Log how frequently it is reached within a predeclared horizon, including failed and expired setups.

Make recognition testable

For every pattern rule, record:

• The swing-selection rule and minimum history.
• The geometry tolerances used to accept the shape.
• When the formation became observable without future candles.
• The entry trigger, cancellation rule, and maximum waiting time.
• The invalidation, target convention, and execution assumptions.

If the strategy keeps adjusting the neckline or target after entry, the original test no longer describes the live process. Version rule changes explicitly rather than rewriting losing examples.

Limits and further reading

Automated recognition is parameter dependent, and different formations often overlap on the same candles. Ten labels on one move are not ten independent pieces of evidence. See CME Group's technical analysis course for charting concepts and backtesting for testing a pattern rule.