Candlestick patterns name price shapes and sequences. They help describe a chart; a name does not establish a probability of profit. Start by measuring bodies and shadows and recording when the completed pattern is known. OHLC values describe the selected price series and interval, rather than the identity, intentions or continuous control of buyers and sellers.

What the shape of a candle actually tells you

O is the interval’s first price, H its highest, L its lowest and C its last. Body length is |C − O|; upper shadow is H − max(O, C), lower shadow is min(O, C) − L and total range is H − L. A rising body has C above O and a falling one C below O; colours depend on chart settings. These formulas do not reveal whether H or L was reached first.

A long body or shadow is a measurement, not proof of conviction or rejection. Identical OHLC can come from different intrabar paths, so a large rising body does not establish uninterrupted buying control. A shadow records an extreme outside the body without naming the traders involved. The guide to candlestick patterns worth knowing first provides terminology; keep interpretation separate from the measured shape.

A doji made from two other candles

A candle’s shape describes its chosen time interval. Changing that interval can give the same history a different label.

Scroll sideways to see the full diagram

A doji made from two other candlesOur two H1 candles are O 100, H 101, L 89, C 90 and O 90, H 102, L 88, C 100. Neither has a zero body. Combining them into H2 gives O 100, H 102, L 88, C 100 and a zero body. All three drawings share a linear price scale; H2 aggregates H1 rather than being a third consecutive observation.The interval changes candle shapeTwo H1One H2889410009–1010–1109–11H1: each body spans 10 unitsH2: O = C = 100 → body 0H2: H = 102 • L = 88

Our own data: Two consecutive fictional H1 candles from one series, without missing observations. First: O 100, H 101, L 89, C 90. Second: O 90, H 102, L 88, C 100. Combine exactly 09:00–10:00 and 10:00–11:00 into 09:00–11:00; keep the time zone and price type unchanged.

Aggregation takes the first O, highest H, lowest L and last C. This gives O 100, H 102, L 88, C 100: a zero body. Neither H1 was a doji under the strict O = C definition. A larger interval does not preserve its component candle labels. OHLC documentation defines the period’s first, highest, lowest and last prices. This is an aggregation example, not a trade signal.

A doji on H2 does not mean each component H1 was a doji.

Single-candle shapes and their preceding context

Hammer and hanging man share a small body near the top, a long lower shadow and little upper shadow; the classic hammer description asks for a lower shadow at least twice the body. The name depends on a preceding decline or rise. Inverted hammer and shooting star instead have a long upper shadow and small body near the low, distinguished by preceding decline or rise. None forecasts a bounce by shape alone. The hammer pattern guide separates geometry from its context.

Ten recognition names; no performance ranking
NameShapeContext in the cited glossary
HammerSmall upper-position body; long lower shadowAfter decline
Hanging manHammer-shaped candleAfter rise
Inverted hammerSmall lower-position body; long upper shadowAfter decline
Shooting starInverted-hammer-shaped candleAfter rise
DojiEqual or nearby O and CNo direction from the body alone
Spinning topSmall bodyAssess relative size
Bullish engulfingRising body contains earlier falling bodyAfter decline
Bearish engulfingFalling body contains earlier rising bodyAfter rise
Morning starLarge falling body; small body gapped lower; rising close deep inside first bodyThree-bar sequence after decline
Evening starLarge rising body; small body gapped higher; falling close deep inside first bodyThree-bar sequence after rise

A doji has equal or nearly equal O and C; a spinning top has a small body. Define “near” and “small” numerically in a scanner. A body at most 5% of its own high–low range is one possible invented rule, not the universal doji standard. Handle a zero range without division by zero. The doji and market indecision guide explains the terminology; equal prices do not prove weakened confidence or a future direction.

Multi-candle sequences and boundary rules

An engulfing candidate compares the second candle’s body with the first, rather than erasing every price in the first candle’s full High–Low range. A bullish second body contains the earlier bearish body under a stated boundary convention; the bearish version mirrors it. The shape alone does not establish a reversal probability. The engulfing guide develops the recognition rule.

The classic morning-star description includes a gap between the first large falling body and the second small body, followed by a rising close well inside the first. Evening star mirrors it. A no-gap detector is a modified rule rather than silently identical. Three white soldiers have three rising candles with progressively higher closes near their highs; three black crows have relatively long falling bodies closing near their lows. More candles do not automatically establish more reliable signals. The morning star and three white soldiers guides develop these different definitions.

Hesitation patterns: harami and half retracements

Harami compares a small second real body with an unusually large preceding body. The second body, rather than necessarily its entire high–low range, lies inside the first. The cited glossary permits either colour for the second candle; stricter scanners may choose another convention. Describe the trend context separately instead of treating a later candle as a guaranteed direction decision. See the harami pattern for explicit rules.

Two additional names use penetration into the previous body. In the cited piercing description, a long falling body after a decline is followed by a lower opening and a rising close more than halfway into that body. Dark cloud cover has a rising first body, a higher opening and a falling close well into it, preferably more than halfway. “Roughly halfway” alone is incomplete, and a partial penetration does not establish weaker performance than engulfing. Specify the gap and penetration convention.

Why no candle is a signal on its own

No verified FX reversal probability is supplied for these candle shapes. A historical stock-study percentage requires its source, sample, definition and outcome horizon before comparison with another market. Specify chart context and confirmation rules, then test the full strategy, including costs, rather than infer an edge from a pattern name.

Define any context filter rather than assume that support, a higher-timeframe trend and a later candle provide independent evidence. Several labels may use the same prices. Adding conditions changes the eligible sample and requires a comparison after costs. The guide to draw support and resistance helps record levels, but no level or extra label multiplies signal quality by definition.

Use fictional EUR/USD data: O = 1.1000, H = 1.1006, L = 1.0980 and C = 1.1005, with a 0.0001 pip. Body = 5 pips, upper shadow = 1, lower shadow = 20 and range = 26; 1 + 5 + 20 = 26. The lower shadow is four times the body. To choose a hammer context in this exercise, define a prior decline as three lower closes: 1.1050, 1.1030, 1.1010. That is our context rule, not a universal trend definition or evidence of a bounce.

A second fictional pair distinguishes body engulfing from full-range engulfing. Candle A: O = 1.1020, C = 1.1000, H = 1.1040, L = 1.0990. Candle B: O = 1.0998, C = 1.1022, H = 1.1030, L = 1.0995. B’s rising body contains A’s falling body, while B’s entire high–low range lies inside A’s range. Both statements are true; preceding context and boundary conventions remain separate.

Two labels can also share one observation. Take 12 fictional warm-up candles O/H/L/C = 100/110/98/105, then 100/101/96/98 and 97/102/85/101. On the final bar the pinned TA-Lib C 0.6.4 engulfing and hammer conditions each give code 100 with default settings. The last body has length 4, lower shadow 12 and upper shadow 1; its body 97–101 contains the previous 98–100 body. The two codes are classifications of one bar, not two independent observations or probabilities. Those functions do not themselves check the preceding trend.

The site’s published comparison of 120 completed Kraken EUR/USD trade-price H1 bars from 21–25 September 2026 reports 83 labels across 59 bars using 18 detectors; 19 bars have multiple labels. These are counts from that saved sample, not a Forex-wide frequency or profit study. Keep one instrument/interval/time identifier per bar and pattern labels as separate columns. Testing trades additionally requires quote-side fills, exits, costs and signal availability; a later confirmation cannot justify an earlier entry at that candle’s best price.

What to do tomorrow

  1. Fix the data and definitions. Record instrument, price basis, interval, time zone, pattern thresholds and context rule before classifying a chosen chart segment.
  2. Measure bodies and shadows. Calculate |C − O|, both shadows and H − L. Check the sum and handle zero-range candles separately.
  3. Record detection time and shared labels. Distinguish a pattern’s bar location from when it becomes known, and keep one observation identifier even if several detectors flag it.
  4. Test complete rules on later data. Keep failures and costs and avoid selecting only attractive examples. A week on demo or repeated labelled shapes does not establish live readiness or profitable execution.

If you want to organise your knowledge of candles within a broader course, a good starting point is the chapter on technical analysis in our library.