Candlestick patterns: which ones survive a backtest, and which don't

Candlestick patterns are taught with a confidence the evidence does not support. Memorise fifty named shapes, the courses imply, and you can read the market. In reality a candlestick pattern is a small, noisy piece of information whose value depends almost entirely on where it appears. Used as a standalone signal, most patterns barely beat a coin flip. Used as a timing tool inside a context, a few of them genuinely help.

What a candlestick pattern actually is

A candle summarises a period of trading into four numbers — open, high, low, close — and its shape encodes who won that period. A long lower wick means sellers pushed price down and buyers fought it back up. A large body means one side dominated. That is all a pattern is: a compact description of a short-term shift in the balance of buying and selling pressure. It is a real piece of information. It is also a tiny one, and a single period of trading is mostly noise.

Why context decides everything

The same pattern means opposite things in different places. A bullish engulfing candle in the middle of a featureless range is noise — it predicts almost nothing. The identical candle at a well-defined support level, after an extended decline, is a meaningful signal: it shows buyers stepping in exactly where you would expect a reversal to begin. The pattern did not change. Its location did, and location is most of the signal. This is why pattern-spotting in isolation fails to backtest — it strips away the context that was carrying the information.

Which patterns tend to hold up

The patterns that survive honest testing share a trait: they describe a clear, large shift in pressure, not a subtle one. Engulfing candles and pin bars (long-wicked rejection candles) encode an unambiguous, visible fight and a winner — and at the right location they carry weight. The elaborate three- and four-candle patterns with exotic names are mostly fragile: rare, loosely defined, and easy to fit to past data. As a rule, the simpler and larger the pressure shift a pattern describes, the better it tends to test.

Defining a pattern objectively

"It looked like a pin bar" cannot be backtested. To test a pattern you must define it as arithmetic on the four prices — for a pin bar, something like: the wick on one side is at least twice the body, the body sits in the far third of the range, and the candle appears after a move of a given size. Precise, repeatable, no eye involved. Then — and only then — gate it with context (a trend filter, a support level) and backtest the whole rule. You will usually find the context does most of the work and the pattern just sharpens the entry timing. That is the honest result, and it is still useful.

Noon Barbari's strategy designer lets you express candle conditions as objective rules on open/high/low/close and combine them with trend and level filters, so a 'pattern' becomes something you can backtest honestly rather than spot by eye after the fact.

Candlestick patterns are not spells and they are not useless. They are small signals that work only with the context that surrounds them. Define the pattern as arithmetic, gate it with context, backtest the pair — and you keep the part that is real.