Dragonfly Doji: A Day Given Back Entirely
A dragonfly doji is a candle whose open and close are both at the top of its range, leaving a long lower wick and effectively no body. It is read as a session where a decline was completely reversed, and its defining feature - the missing body - is easier to produce on illiquid instruments.
How it works
A dragonfly doji opens and closes at essentially the same price, and that price is at the top of the bar’s range. Everything below is wick. On a chart it looks like a T.
“Doji” means the body is negligible relative to the range. A common working threshold — and the one used for the counts on this page — is a body of 10% of the range or less. Dragonfly adds the second condition: that near-zero body sits at the top, with the lower wick covering most of the bar.
So it is a hammer taken to its extreme. Same story, more of it: price fell through the session and returned to exactly where it started.
The story, and what makes it fragile
The reading is total rejection. Whatever the sellers achieved during the session was undone by the close, and price finished where it opened. Told that way it sounds like the strongest possible version of a rejection candle.
But the defining feature is an absence — the missing body — and absences are easy to manufacture. On an instrument where the minimum price increment is large relative to typical movement, open and close land on the same tick regularly, and every such bar with a lower wick becomes a dragonfly.
Which means the pattern’s frequency is partly a property of the instrument’s tick size and liquidity, not of participant behaviour. On a liquid future with a fine tick, a true zero-body bar is rare and meaningful. On a penny stock, it is Tuesday.
Volume is the check that separates the two cases. A dragonfly on heavy participation means a genuine round trip happened with real size on both sides. On light volume it means almost nothing traded, which is how a bar ends up with no body in the first place.
In practice: the count and the comparison
On this site’s shared 576-bar history there are 75 doji of any kind and 26 dragonfly doji — 4.5 per hundred bars. About a third of all doji here have their body at the top.
Of those 26, the next bar closed higher 54% of the time.
Across all 575 transitions in the same data the next close was higher 51% of the time. Three points of difference, on 26 events.
Twenty-six occurrences cannot support a claim about anything. Move two of them the other way and the edge is gone; move two more and it inverts. That is not a criticism of this particular count — it is the condition every candlestick statistic is in, including the ones quoted with far more confidence elsewhere, because rare patterns are rare and there is no way around the arithmetic.
Zoom out and the shape is destroyed or created at will. A weekly dragonfly requires the week’s open and close to match; a daily one requires the day’s. The same price history produces the pattern at one aggregation and not at another, which is a strong hint that the pattern is a property of the chart settings.
And after a gap, the open is a price nobody traded at yesterday. A dragonfly whose open sits at a gap level is not a session returning to where it started — it is a session returning to a level that was itself created by the market being closed.
Each one traded costs 2% of a typical bar’s range in round-trip costs on this history — 26 payments against a three-point difference that a sample of 26 cannot establish.
What a dragonfly doji is not
It is not a gravestone doji. That is the same idea inverted: body at the bottom, wick above. They are mirror images with opposite readings.
It is not a hammer. A hammer has a small but real body. The distinction is a threshold somebody chose, and bars near that threshold get sorted into different pattern names on a difference of one tick.
It is not indecision. “Doji means indecision” is the standard line, and it describes the open and close matching rather than anything about participants. A session with a violent decline and a full recovery is not indecisive; it is eventful and symmetrical.
And it is not more reliable for being dramatic. The visual impact of a long T on a chart has no relationship to the size of whatever statistical effect it has, and the impressiveness is a large part of why it gets traded.
When it fails
In a range they appear at both edges. Price probes the floor and returns: dragonfly. Price probes the ceiling and returns, and depending on where the open falls you can get one there too. The pattern occurs most often in the regime where a reversal signal is least useful, because everything reverses.
The second failure is the threshold. “Body under 10% of range” is one convention; others use 5%, or require an exact match of open and close. The count of dragonfly doji on any chart changes substantially with that choice, which means published frequencies are not comparable.
A third is the illiquidity trap. The pattern is most common on instruments where it means least, because a missing body is easiest where little trades.
A fourth is placing the stop under the wick. The low of a dragonfly is, by construction, an extreme that price visited briefly and left. It is the most obvious price on the chart, and obvious prices are where resting stop orders cluster.
And a fifth is expecting the reversal to be immediate. Even taken at face value, the pattern says something about the session that just ended. It carries no timing information about what comes next, and “it eventually reversed” is a claim that can be made about any bar on any chart.
The original data
26 dragonfly doji in 576 bars — a third of the 75 doji of any kind in the same series — followed by a
higher close 54% of the time, against a 51% base rate across all 575 transitions. The definition
thresholds and the counts sit in research/series-measurements.json, generated by
site/measure_series.py.
The 75-to-26 split is the figure worth carrying away. Two thirds of the doji on this chart are not dragonflies, which means the doji family is broad and the specific sub-pattern is uncommon — and uncommon patterns produce small samples, and small samples produce statistics nobody should act on. Before trusting any candlestick figure, ask for the occurrence count first and the percentage second. That ordering disqualifies most of what is published about this pattern, and it should.
Related
Candlestick patterns is the parent page for the whole naming system. Hammer candle is the version with a real body and the fuller comparison. And candlesticks covers what the four prices in any bar actually record.
The dragonfly looks like the most dramatic candle on the chart, and that is most of its appeal. What changed how I read it was noticing how often the tiny body was simply the smallest increment the instrument could move in - the drama was in the tick size, not in the session.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.