Double candlestick patterns: body relationships and context

Use the appropriate boundary for each family

For each candle, the real body runs from the lower to the higher of open and close. The full range runs from low to high. Confusing these intervals changes which bars qualify. Record whether equality is allowed at each boundary, and define small or long bodies numerically when using a detector.

Worked example: body engulfing without range engulfing

Consider hypothetical candle A with open 105, high 108, low 100 and close 102, followed by candle B with open 101, high 107, low 100.5 and close 106. B covers the full A body from 102 to 105. Its high is lower and its low is higher than A, however, so B is also inside the first high-to-low range. These labels measure different relationships and can coexist.

• Assume a prior decline for a bullish reversal hypothesis. Without that assumption, the two bars establish only the geometry.
• If B instead opens at 103 and closes at 104, its body is inside A: that is a harami-shaped contraction, not engulfing. Body size and any color rule still need to be specified.
• If the opening gap of a piercing pattern must be below A low of 100, an open of 101 fails that gap test even though it is below A close of 102.
• All numbers are explanatory prices, not an identified instrument, trade or backtest.

Define context without changing it after the outcome

A bullish formation after a local decline can occur inside a longer uptrend. That can be a reversal of the local decline and continuation of the larger move at the same time. Record both horizons instead of declaring the shape invalid because a broader chart points upward. A fixed five-bar, ten-bar or moving-average trend filter is a choice to test, not a universal requirement.

• Use only pre-pattern prices to classify the prior move. Keep range-bound and unclassified cases as separate groups.
• A candle represents the selected interval, which can span minutes, hours or days. It is not necessarily one exchange session.
• Fix regular or extended session settings and ordinary versus synthetic chart data before comparing detectors.

Separate recognition from an entry experiment

In the worked engulfing example, the pair high is 108 and low is 100. One illustrative confirmation rule waits for the next close above 108 without a prior break below 100. A next high of 109 followed by a close of 107 does not confirm that rule. Expire the setup after that next bar if it has not qualified; any longer deadline should be a separately named variant.

• Recognition occurs when B closes; later confirmation belongs on its later bar. Neither event fixes an order fill price.
• A strategy must also define entry order, stop, target or time exit, size, fees and slippage. A 1:2 target specifies payoff distance, not observed return.
• Compare optional volume and RSI filters with the unfiltered sample. Higher volume does not identify institutional traders, and a conventional RSI boundary is not proof of a reversal.
• Keep failures and expired candidates. If one bar crosses both confirmation and cancellation boundaries, choose and disclose a treatment for the unknown intrabar ordering.

Compare pattern families on the same basis

To compare a single-candle and a two-candle strategy, use the same instruments, dates, costs, exit logic and position sizing. Record when each signal becomes observable. Report trade count, net outcome and drawdown; do not rank detectors by body size or by how many bars their names contain.

• Freeze definitions before selecting a development period, then evaluate on a later period excluded from parameter selection.
• Two overlapping labels can describe the same price movement. Decide whether they produce one candidate or separate candidates before counting results.
• No best timeframe or fixed accuracy is established by this comparison guide.