Skip to main content

Single-Candle Patterns (Doji, Hammer, Marubozu)

Updated Jun 23, 2026 at 8:47pm

Research Draft Medium 1,377 words

Single-candle patterns are interpretations of one bar's open-high-low-close geometry — the size of the real body relative to its upper and lower shadows — as a snapshot of the intra-period tug-of-war between buyers and sellers. A long body means one side dominated the whole period; a tiny body with long wicks means the period closed near where it opened after a fight. The core tension of this family is that the shape is necessary but not sufficient: the same silhouette is bullish, bearish, or meaningless depending entirely on the prior trend in which it appears, and even with correct context the measured standalone edge of these signals is weak (see Standing & evidence). Treat them as flags that something may be shifting at this bar, requiring confirmation — not as standalone trade triggers. Multi-bar combinations (engulfing, harami, morning/evening star, three-candle patterns) are covered in the sibling Two- and Three-candle nodes; swing entry/stop/target mechanics live in the Swing Trading branch.

How they're formed

Each pattern is defined by the relationship between the real body (open-to-close range) and the shadows/wicks (the extremes beyond the body). Standard definitions (StockCharts ChartSchool):

  • Marubozu — a candle with no shadows (or virtually none); the open and close are the high and low. A white/bullish marubozu opens at the low and closes at the high (buyers controlled the entire period); a black/bearish marubozu opens at the high and closes at the low. Signals one-sided conviction. Context-independent in meaning (it always shows control) but its predictive use still depends on location.
  • Doji — open and close are virtually equal, producing a near-zero body (a cross, plus, or "T"). Signals indecision. Variants by shadow geometry:
- Long-legged doji — long upper and lower shadows of roughly equal length; maximum indecision (price ranged far both ways, closed at the middle). - Dragonfly doji — open/high/close cluster at the top, long lower shadow, looks like a "T". Sellers drove price down intraperiod; buyers recovered it all. - Gravestone doji — open/low/close cluster at the bottom, long upper shadow, an inverted "T". Buyers pushed up; sellers recovered it all.

  • Hammer — small real body near the top of the range, long lower shadow at least 2–3× the body, little/no upper shadow. Must appear after a decline. Bullish reversal candidate (sellers pushed lower, buyers reclaimed).
  • Hanging Man — identical shape to the hammer (small body, long lower shadow) but appearing after an advance. Bearish reversal candidate. Same silhouette, opposite context, opposite implication — the cleanest illustration of context-dependence.
  • Shooting Star — small body near the bottom of the range, long upper shadow ≥ 2× the body, little/no lower shadow, after an advance. Bearish reversal candidate.
  • Inverted Hammer — identical shape to the shooting star (small body, long upper shadow) but appearing after a decline. Bullish reversal candidate.

How they're read

Two dimensions decode every single-candle signal: (1) the geometry tells you what happened inside the bar (who pushed, who recovered); (2) the prior trend tells you what it means. A long lower wick after a downtrend = sellers exhausted (potential bottom, hammer/dragonfly); the same long lower wick mid-range or after an uptrend (hanging man) = warning of late-stage buying. A doji after a strong uptrend says momentum is stalling; a doji in a sideways chop says nothing. The shape mid-trend or mid-range carries essentially no signal — these are reversal-context tools.

Because the standalone reliability is low, the canonical reading rule is wait for confirmation on the next bar: a hammer is confirmed by a gap-up or strong white candle (ideally on rising volume); a shooting star/hanging man by a gap-down or strong black candle. The candle marks a location of interest (potential support/resistance); confirmation supplies the trigger.

How they're used in practice

The recognized, style-agnostic uses are: (1) reversal alerts at support/resistance — a hammer printing into a prior support zone or a gravestone doji into resistance flags a possible turn, which the trader then confirms and trades off the structure, not the candle alone; (2) exhaustion/indecision signals within a trend — a doji or long-legged doji after an extended run warns a trend-follower that momentum is fading, often used to tighten stops rather than enter; (3) conviction confirmation — a marubozu in the breakout direction corroborates a breakout's strength (full-period one-sided control), used as supporting evidence alongside volume; (4) as one input in a confluence stack, never alone — combined with trend, volume, S/R levels, and indicators. Across all uses the operative discipline is candle proposes, confirmation + context disposes. Exact swing entry/stop/target sizing belongs to the Swing Trading branch.

Standing & evidence

This is the section traders most need and most skip. The measured standalone edge of single candles is weak, and this is well-documented:

  • Bulkowski (Encyclopedia of Candlestick Charts; thepatternsite.com) backtested each pattern. The hammer acts as a bullish reversal ~60% of the time — which he explicitly calls "not bad, but also not far from random (50%)" — and ranks 65th of 103 candle types on overall performance. Most doji variants are worse and hover at coin-flip: the dragonfly doji acts as a reversal ~50% ("random"), the gravestone doji as a bearish reversal ~51% ("random"), the southern doji ~52% ("near random"), and the northern doji continues ~51% ("about random"). A few specific gapping variants edge higher (gapping-up doji ~57% bearish reversal). Attribution matters: these are Bulkowski's tested rates, not laws.
  • Academic skepticism: Marshall, Young & Rose (2006), "Candlestick technical trading strategies: Can they create value for investors?" (Journal of Banking & Finance 30(8)), tested candlestick signals on DJIA component stocks 1992–2002 using a bootstrap that generates random OHLC series. They found candlestick strategies do not produce statistically significant excess returns in large-cap US stocks — consistent with informational efficiency. (Note the scope: large-cap US equities, that period; later studies on other markets/assets report mixed, sometimes positive results, so this is evidence against a universal edge, not proof candles never work anywhere.)

The honest synthesis: single candles are real, named, widely-taught descriptions of price action, but as standalone predictive triggers their tested win rates sit near 50% and their measured edge in rigorous studies is weak-to-absent. Their value, where it exists, is contextual and confirmatory.

Strengths & limitations

Strengths: fast, universal, intuitive read of intra-bar control; useful as exhaustion/indecision alerts at known S/R; the marubozu is an honest conviction marker.

Limitations / failure modes: (1) context-blindness is the #1 misuse — calling a "hammer" mid-range or a "doji" in chop where the shape has no reversal meaning; the hammer vs. hanging man (same shape, opposite trend, opposite meaning) is the canonical trap. (2) Standalone trading — entering on the candle alone, ignoring the near-random base rates. (3) Timeframe/liquidity dependence — on thin or low-timeframe data a single tick can manufacture or erase a wick, so the "pattern" is noise; higher timeframes and liquid instruments are more meaningful. (4) Definitional looseness — exactly how long a shadow or how small a body qualifies varies by source, so two analysts can disagree on whether a bar "is" a hammer. (5) Survivorship in the folklore — vivid winning examples are remembered; the ~40-50% that fail are not.

System relevance

Within Delvantic, single-candle recognition is a low-weight contextual input, not a trigger. The Augustus trade-setup agent should treat a single candle as corroborating evidence only inside a confluence stack (trend + S/R + volume + the multi-candle and swing nodes) and should down-weight or ignore any single-candle signal that lacks prior-trend context or next-bar confirmation, reflecting the near-random base rates above. Cross-links: the Two-/Three-candle sibling nodes (combinations with higher tested reliability) and the Swing Trading branch (operational entry/stop/target). Hard caveat to carry downstream: do not let a named candle shape inflate conviction on its own — its measured standalone edge is weak.

Sources