Candlestick names describe one OHLC bar or relationships between several bars. Start with the prices and a numerical definition rather than a ranking of effectiveness. This guide compares real bodies, shadows and full ranges, then shows why several labels can describe the same observation without becoming independent confirmations.

A green candle can close below the previous close

The body measures movement from this candle's open to its close. Comparing with the previous close requires another price.

Scroll sideways to see the full diagram

Body colour and closing change are different comparisonsTwo identical rising candles O100 H103 L99 C102 have previous closes of 105 in A and 98 in B. Close-to-close changes are −3 and +4 respectively.GREEN BODY ≠ HIGHER CLOSEA: previous Close 105C 102O 100C − previous Close: −3B: previous Close 98C 102O 100C − previous Close: +4Both bodies: 102 − 100 = +2.A: gap −5 plus body +2 totals −3.B: gap +2 plus body +2 totals +4.Colour compares C with O, not prior C.

Fictional arbitrary price scale, no specified market or trade result. Both panels use O=100, H=103, L=99, C=102, with green for C>O. Previous Close is 105 in A and 98 in B. The segment from previous Close to O is an arithmetic gap here, without an assumed cause or frequency. The document CME Group — Chart Types confirms that open and close form the body and high and low determine the wicks.

The same +2 body coexists with a −3 or +4 close-to-close change. The difference comes from the open relative to the previous close.

Journal C−O and C−previous Close separately. A rising candle describes the first measure. When assessing the closing-to-closing change, state the second explicitly; green alone does not establish it.

The pin bar — a candle that rejects a price

A pin bar is an informal name for a small real body and a dominant shadow on one side. Define the thresholds before classifying cases: for example, lower shadow at least twice the body and body no more than 30% of the range, with a separate upper-shadow condition. These are chosen measurement rules, not evidence of profitability. Candle colour compares its close with its own open.

For ordinary OHLC data, range = H − L, body = |C − O|, upper shadow = H − max(O,C) and lower shadow = min(O,C) − L. In a fictional EUR/USD bar, O = 1.1000, H = 1.1010, L = 1.0960 and C = 1.1005. At 0.0001 per pip, range is 50 pips, body 5, upper shadow 5 and lower shadow 40. The body is 10% of range and the lower shadow eight times the body. This records shape, not which participants acted or the order in which H and L occurred.

If a rule also uses support or resistance or a Fibonacci retracement, define that condition and when it was known. Its presence does not automatically prove an advantage; the measured bar exists whether or not it meets the chosen context filter.

Pin bar — an example rule to specify
Shape of the candleChosen example: dominant shadow ≥ 2 × body, body ≤ 30% of range; define the other shadow
Context of the levelIf a level filter is used, save its definition before inspecting outcomes
Entry ruleFor a closing rule, wait for the final price; define trade direction separately from colour
Stop lossSpecify order conditions and buffer; a stop level does not guarantee the fill or loss
TargetDefine the exit anchor and projected reward-to-risk; neither proves positive expectancy
When to passWhen the stated shape/context rule fails or position constraints do not permit the planned order

Engulfing — the candle that swallows the one before it

An engulfing pattern forms when the body of a candle completely covers the body of the previous one of the opposite colour. In the bullish version a small down candle is followed by a large up candle that covers its predecessor from above and below. The bearish version looks the other way round.

Engulfing compares real bodies, not necessarily the complete High–Low ranges. In a fictional pair of EUR/USD candles, the first is O 1.1020, H 1.1040, L 1.0980, C 1.1000; the second is O 1.0995, H 1.1030, L 1.0990, C 1.1025. The second body 1.0995–1.1025 contains the first 1.1000–1.1020, yet its full range is inside the first. Both labels can apply. Specify how equality is treated; the engulfing pattern as a reversal signal guide develops those boundary comparisons.

The doji — a sign of indecision, not of reversal

A doji has open and close near one another; record a tolerance rather than calling every small body a doji. It does not by itself select the next price direction. A rule waiting for another completed candle has a later information time, not guaranteed confirmation of who gained control. See the doji as a pattern of indecision guide for definitions.

  • Define resistance and any subsequent-candle condition as part of one testable short-rule variant.
  • Define support and any subsequent-candle condition separately for a long-rule variant.
  • If a context condition fails, record no signal under that rule; do not infer that the OHLC observation contains no information.

Hammer and shooting star — the pin bar in the context of a trend

A hammer has a small body near the top and a long lower shadow after a decline; similar geometry after an advance is called a hanging man. A shooting star has a long upper shadow after an advance; a similar shape after a decline is an inverted hammer. State both geometry and the prior-trend rule, rather than treating every long wick as the same reversal setup.

Those names are conventions for shape and context, not measurements of control. The same OHLC can arise through different intrabar paths. If a final Close is required, identify the completed bar; if using an unfinished bar, describe a different rule whose values can still change.

The inside bar — compression before a breakout

An inside bar has its High–Low range within the preceding bar’s range. Define strict versus inclusive boundaries. It records a narrower range in that comparison, not a forecast of breakout timing or a full volatility estimate. It differs from harami, which compares bodies, and can coexist with real-body engulfing in the earlier example.

Inside bar — keep geometry separate from an order rule
Identificationcompare both High and Low with the preceding bar; specify equality
Upside variantif using a buy stop, define trigger price, buffer, expiry and fill assumptions
Downside variantif using a sell stop, define the corresponding rules separately
Protectiona chosen stop level is a condition, not a guaranteed execution price
Exitany range multiple is a model parameter, not a promised destination

A breakout system needs entry, exit and order-handling rules beyond the inside-bar label. If placing two conditional orders, specify whether and when the other is cancelled. Preserve failed breaks and no-fills in the test rather than claiming a particular trend context guarantees successful continuation.

Three white soldiers and three black crows

Three white soldiers conventionally comprise three consecutive strong bullish candles with successively higher closes, typically opening within the preceding real body and closing near their highs after a decline. Three black crows are the bearish counterpart after an advance, with successively lower closes. They are conventionally reversal patterns in the relevant prior-trend context; shape alone does not guarantee the next move.

A three-candle shape does not by itself define an entry or prove that it is too late. Record the time the third close became available and assess any subsequent order at executable prices. Neither candle count nor increasingly high or low closes establishes the future direction or a fixed stop distance.

Morning star and evening star — a reversal in three candles

The morning star is conventionally a potential reversal after a decline. In the stated body-gap convention, check three distinct conditions rather than three candles of any shape.

  1. The first is a large down candle, a continuation of the move so far.
  2. The second has a small real body separated below the first body by a gap; state the numerical gap and small-body definition.
  3. The third is bullish and closes above the first real body’s midpoint in this chosen convention. A detector may use a different penetration parameter, so name its implementation.

The evening star reverses the geometry after an advance. A version omitting a required body gap is a modified rule and should be labelled accordingly. Three bars do not prove higher predictive credibility than one. For example, TA-Lib’s morning-star implementation uses configurable body lengths and penetration and does not itself check the prior trend; do not equate a returned label with a complete trading setup.

Practical rules for using the patterns

  1. Name the data interval. Record the instrument, feed, time zone and price type. No evidence here makes every M5 pattern noise or every H4 pattern reliable. Compare specified rules on data not used to choose their parameters.
  2. Define any context filter. Save the level and trend rule before examining later outcomes. A filter that fails means no signal under that rule, not an absence of measured candle data.
  3. Specify additional conditions. Candle patterns, levels and RSI are often derived from the same prices, so they are not automatically independent evidence. Test whether the combined rule improves results after costs rather than assume it moves the odds away from a coin flip.
  4. Respect signal availability. A rule using the final Close becomes available at the end of the bar. Its unfinished values can change. An intrabar rule is a separate protocol; the shape does not have to disappear to distinguish the two.
  5. Separate the stop condition from realised loss. Specify the trigger price, buffer, order terms and position volume. Ordinary stops can execute at a different price; the outer wick does not guarantee a loss bound. Include commissions and funding, and use Bid/Ask prices without deducting the spread twice.

Practical takeaway: A candle's body colour compares its own close and open, not its close with the previous candle's close.

What to do before you open a position

The five steps below take a few minutes. If any of them comes out as a no, it is better to let the opportunity go than to hunt for arguments in its favour.

  1. Record the timeframe and the exact candle interval. Compare another interval only as a separately defined dataset, rather than dismissing a bar solely because it is below H4.
  2. Measure range, real body and both shadows. Record threshold and equality rules. In the earlier model, changing only Close from 1.1005 to 1.0980 makes the body 20 pips, upper shadow 10 and lower shadow 20. The lower-shadow/body ratio becomes 1, so the chosen two-times condition fails.
  3. If using levels, moving averages or other filters, freeze their definition before assessing the outcome. Candle patterns and price-derived indicators can reuse the same history; agreement is not automatically an independent vote.
  4. Check scheduled and simultaneous news when designing observation windows. A calendar does not quantify the likelihood of an invalidation, and a pattern label does not guarantee order execution during a release.
  5. Calculate the stop loss and the position size before the entry, never after it. The distance to the point where the idea stops making sense is an input, and it is what sets the size of the position, not the other way round.

For a broader survey of the patterns alongside indicators, chart formations and multi-timeframe context, see the technical analysis section on ForexMechanics.