A doji with a long lower wick and no upper wick, after a fall.
Try it
The rules, live
Bend the candles and watch each rule pass or fail. The verdict comes from the same code that counted the stats below.
Drag the handles (H high, L low, O open, C close), or Tab to one and use the arrow keys.
This is a dragonfly doji: all 3 rules pass.
BeforeComes after a fall: the day before, the stock closed lower than 5 trading days earlier and below its 10-day average.
CandleOpen and close are almost equal: the body is at most 10% of the day's range.
CandleOpens and closes at or near the high: the upper wick is at most 10% of the day's range.
These are the exact rules the stats below were counted with, and the ones ta.candlePattern uses in a script. Textbooks differ in the details; every threshold here is fixed and public.
Why traders watch it
What it says about buyers and sellers
Sellers drove the price well down during the day, then buyers bought it all the way back to where it opened. After a fall, it shows sellers losing their grip, though the day itself ended level.
It only counts after a fall: the same shape elsewhere means something else, or nothing.
Honest numbers
What it actually did next on NSE stocks
Every case our rules found in the data, compared with buying on any day.
Verdict · Beat costs, not steadily
After a dragonfly doji, stocks rose 5 trading days later 49.2% of the time, against 46.4% on any day, and the average move up was 0.41% bigger than on any day: more than the 0.30% cost of a round trip. But it was not steady: better than any day in only 8 of 12 years with enough cases, so it may be luck.
Higher 5 days later
49.2%
Any day: 46.4%
Average move
+0.46%
Any day: +0.05%
Typical move (median)
−0.04%
Any day: −0.38%
Cases
2,074
1 Jan 2015 to 25 Sept 2026
Measured from the next day's open (you only know a pattern at the close) to the close 5 trading days after the pattern. The median is the middle case: half did better, half worse. It is below the average because a few big gains pull the average up.
The edge against the cost of trading it
Extra average move up, vs any day
+0.41%
Cost of a round trip
0.30%
On average the extra move covered the cost of a round trip (₹298 on ₹1,00,000), before tax on gains. Averages hide many losing trades.
Year by year: how often higher 5 days later
Dragonfly dojiAny dayBeat any day in 8 of 12 years
Point at a year for its numbers. Faded years had fewer than 30 cases and don't count.
How these numbers were made
Daily candles of 2,032 NSE stocks from 1 Jan 2015 to 25 Sept 2026 (companies with filed financials, so no ETFs or index funds). The rules are the ones shown above, exactly.
A day counts only if the stock closed at ₹20 or more and its median daily traded value over the last 20 sessions was at least ₹1 crore, so you could actually have traded it. "Any day" uses the same days, and is weighted to the same years as the pattern.
Corporate actions: these prices are adjusted for splits and bonuses. Evidence: RELIANCE's 1:1 bonus (ex-date 28 Oct 2024) shows no overnight halving in this data: ₹1,327.85 close on 25 Oct, ₹1,337 open on 28 Oct. Its 2017 bonus (ex-date 7 Sep 2017) is the same: ₹392.10 close, ₹392.25 open. As a safety net, any stretch with a jump of more than 35% from one close to the next open or close, or a gap of more than 15 days between sessions, was left out (2,634 days).
"100 most traded" are the 100 stocks with the highest median daily traded value over their last 250 sessions: mostly large, well-known companies.
Only stocks listed today are included. Companies that were delisted or went bust are missing, which flatters every number a little (survivorship bias). Results on the same day across many stocks move together, so the real uncertainty is larger than the case counts suggest; that is why the year-by-year view matters.
Costs: a ₹1,00,000 delivery round trip costs about ₹119 to buy and ₹119 to sell (taxes, exchange and depository charges, no brokerage), plus 3 basis points of slippage each way (₹60): ₹298, or 0.30%. Rates are indicative.
Built 27 Oct 2026. This is history, not a forecast.
Real charts
Real examples
Look at the pattern first, decide what you expect, then reveal what came next.
PERSISTENT: dragonfly doji completed on 19 Jan 2016.
These 6 examples come from the 100 most traded stocks and were picked because the shape is clear, not because of what happened next. Prices are adjusted for later splits and bonuses, so they can differ from what was quoted on the day.
Scripting
Use it in a script
In a strategy script, ta.candlePattern(bars, "dragonfly-doji") gives one true or false per candle: true on the day the pattern completes, by exactly the rules above.
function run(bars, ta, helpers) {
// Buy after a dragonfly doji, sell 5 trading days later.
// The signal goes on the day the pattern completes; the backtest
// fills it at the next day's open.
const hit = ta.candlePattern(bars, "dragonfly-doji");
const signals = [];
let sellOn = -1;
for (let i = 0; i < bars.length; i++) {
if (sellOn === -1 && hit[i]) {
signals.push(helpers.buy(i, "dragonfly doji"));
sellOn = i + 5;
} else if (i === sellOn) {
signals.push(helpers.sell(i, "5 days later"));
sellOn = -1;
}
}
return signals;
}