Japanese candlestick patterns: read the prices before the label

Read a candle without inferring its entire path

On ordinary OHLC candles, the body spans open to close and the shadows extend to the high and low. Candle colors are display settings: an up body closes above its open, while a down body closes below it. Hollow-candle and previous-close color settings can use additional comparisons, so inspect the chart legend rather than assuming a color means the same thing everywhere.

• A hypothetical open of 100, high of 106, low of 98 and close of 104 gives a body of 4, upper shadow of 2 and lower shadow of 2. The total range is 8.
• OHLC does not tell you whether 106 occurred before or after 98. Different price paths can produce the same candle.
• Even a bullish marubozu opening at its low and closing at its high can contain pullbacks within the bar. No wick does not mean uninterrupted buying.
• Use ordinary candles for rules about traded OHLC. Heikin-Ashi transforms those prices and is a different input to a pattern detector.

Single-candle shapes need explicit tolerances

Terms such as small body, long shadow and nearly equal open and close need numeric thresholds for repeatable classification. Decide whether equality is included and what to do with a zero-range bar before testing. The brief descriptions below identify shapes; the linked tutorials provide full example rules.

Two and three candles measure different relationships

Engulfing patterns compare real bodies, not necessarily the full ranges. A harami instead places a small second body inside a larger first body. Morning and evening stars use three bars, including a small middle body and a third close penetrating the first body. Record gap and penetration conventions when comparing detectors.

• Bullish engulfing: an up body covers the preceding down body after a decline. Bearish engulfing reverses the direction after an advance.
• Morning star: a down body, small body gapped lower and recovery candle; evening star reverses those directions. Specify whether the third close must pass the first body midpoint.
• A body gap and a gap between the full ranges are different requirements. A no-gap adaptation should be named separately.
• An extra candle supplies another observation but does not prove a higher win rate. Count overlapping labels consistently if several patterns describe the same move.

Keep formation, confirmation and a trade separate

A closed candle establishes its final OHLC for chart-timeframe measurements. A later close above a bullish pattern high is an optional confirmation rule, not evidence available when the pattern first formed. A confirmed setup can still reverse. Define a deadline and a cancellation boundary before inspecting later bars.

• For an illustrative test, allow only the next bar to confirm a bullish pattern by closing above the full pattern high without trading below its low. Expire it otherwise.
• A high above the boundary followed by a close back inside fails that closing rule, even if an intrabar stop-entry strategy would already have entered.
• Specify actual entry order, exits, position size, costs and slippage. A stop order is not a guarantee of an exact loss limit.
• Volume, RSI or MACD conditions are optional filters to compare on the same sample. Volume alone does not identify institutional participation.

Choose chart settings and evaluate a complete rule

The same time period can produce different candles when the exchange feed, session or chart interval changes. There is no universally best timeframe established here. Select the intended market and interval, keep the definition fixed and include every qualifying candidate rather than only visually attractive reversals.

• Record exchange, symbol, interval, regular or extended session and ordinary versus synthetic candles.
• Separate candidate count, confirmed count and executed trades. A trade CSV cannot show candidates that never entered.
• Compare rules using the same fees, exit method and date ranges. Reserve a later evaluation period after choosing inputs.
• A chart pattern description is not a live scan, an order instruction or measured strategy performance.