I lost 500 dollars on a dragonfly doji once, and the candle was perfect. It printed at the bottom of a three-day fall on EURUSD, a clean T shape with a long tail underneath and no body to speak of, and I bought it on the close with a full position, certain the bottom was in. The next candle opened lower, drifted down through the dragonfly low, and stopped me out before lunch. The candle had not lied. I had traded it backwards. A dragonfly doji is not a buy signal on its own. It is a warning that sellers pushed price a long way down inside one candle and buyers dragged it all the way back, and the trade only exists if the next candle proves the buyers were not finished. Once I stopped buying the T shape and started waiting for the confirmation close above it, the same candle became one of the calmest bottom trades on my funded account.
In this article
- What a dragonfly doji actually is
- The four measurements that decide if the candle counts
- How the dragonfly differs from its neighbours
- Why the next candle matters more than the doji itself
- My entry, stop and targets on one worked example
- The sizing maths under a 5 percent daily loss limit
- I tested it: twelve trades, same rules, no exceptions
- When the dragonfly fails, and how I spot it early
- The session routine I run it on
- How it fits funded rules: drawdown, consistency and patience
- Frequently asked questions
That distinction is the whole article. A dragonfly doji looks dramatic, which is exactly why it is dangerous on a funded account: drama invites size, and size without confirmation invites a breach. Everything below runs on one account model, the same one this site uses throughout: a 100,000 dollar account, a firm daily loss limit of 5 percent, which is 5,000 dollars, and my own rule of 500 dollars of risk per trade, which is 0.5 percent of the account and 10 percent of the daily limit. Five losing trades in a row cost 2,500 dollars and end my day by my own rule, long before the firm limit is in sight. If you trade a different account size, keep the percentages, not the dollar amounts: risk 0.5 percent per trade, never more, whatever your balance.
It is also worth saying what this article is not. It is not a second version of the hanging man, the morning star, or the engulfing pattern, all of which this site already covers. The dragonfly is a single candle, not a three-candle star and not a two-candle engulfing, and its body is the point: a dragonfly has effectively no body at all. That one difference changes the entry, the stop, and the failure modes, and the comparison table below spells it out so you never trade one pattern with another pattern’s rules.
What a dragonfly doji actually is
A dragonfly doji is a single candle shaped like a capital T. The open, the close and the high all sit at or very near the same price at the top of the candle’s range, and a long lower shadow hangs below them. During the session, sellers drove price substantially lower, and by the close buyers had recovered every tick of the decline. After a downtrend, that reads as a demand response at lower prices and is watched as a potential bottoming signal. Before a downtrend, it reads as nothing at all: a T shape in a sideways market is just a candle that went down and came back, which happens all day.
The story inside the candle is what makes it tradeable, and it is worth slowing down for. Sellers were in control at the open. They pushed price down, hard, and for a while the session looked like every other leg of the downtrend. Then buyers absorbed the selling, not at a slightly higher price but all the way back to the open, and the candle closed where it started. Nobody who sold the low is sitting on a profit, and nobody who waited gets a cheap re-entry at the low. That is why dragonflies so often lead to fast follow-through when they work, and also why they punish hesitation when they fail: if you wait for a deep pullback to the dragonfly low, you usually wait forever, because the low itself was the rejection.
For the textbook anatomy of the pattern itself, the Warrior Trading dragonfly doji definition lays out the same T shape clearly: high, open and close all at the same level, with a long lower shadow beneath. What follows is how I turn that shape into funded-account rules, because the textbook does not have a daily loss limit and my account does.
The four measurements that decide if the candle counts
I measure every candidate candle before I even look at the trend around it. If any measurement fails, the candle is not a dragonfly doji, no matter how dramatic the tail looks on the chart. Here is the worked candle from the trade that taught me the pattern, an H1 EURUSD candle at support after a three-day fall:
| Measurement | Rule | Worked candle | Pass or fail |
|---|---|---|---|
| Body (open to close) | No more than 10 percent of the total range | Open 1.0860, close 1.0860, body 0 pips | Pass |
| Upper shadow | No more than 10 percent of the total range | High 1.0862, upper shadow 2 pips | Pass |
| Lower shadow | At least 70 percent of the total range | Low 1.0800, lower shadow 60 pips of a 62-pip range | Pass |
| Prior trend | A clear downtrend into the candle, not chop | Fall from 1.1020 to 1.0850 over three days | Pass |
The body rule is the one most traders fudge, and it is the one that matters most. A candle with a 20-pip body and a 60-pip tail is not a dragonfly doji; it is a hammer, and hammers have different confirmation rules and different stops. A dragonfly has no body worth naming. If you can see a body without zooming in, call it something else. The upper-shadow rule does the same job at the other end: a candle with a long wick on both sides is a long-legged doji, indecision in both directions, not a one-sided rejection. My extra rule, the one that keeps funded-account trades honest, is that the tail must be at least 70 percent of the range. A 30-pip tail on a 100-pip candle is ordinary volatility, not a rejection.
Location matters as much as shape, and it is the second half of the measurement. The dragonfly must print at a level sellers have a reason to lose at: a prior swing low, a demand zone, or a round number that has held before. My worked candle printed at 1.0800, a prior low that had rejected price twice before, after the fall from 1.1020. A perfect T shape in the middle of nowhere is worth nothing, because there is no level for buyers to defend and no reason for sellers to stop. I check location on the higher timeframe before I drop to the entry timeframe, the same habit the trend pullback strategy uses: the level is what turns a candle into a trade, and a candle without a level is just a shape on a screen.
How the dragonfly differs from its neighbours
This site already covers several candles that look like the dragonfly at a glance, and confusing them is expensive, because each one has a different invalidation level. The table below is the check I run before I label any candle, and it is the reason this article exists as its own strategy rather than a footnote to another pattern:
| Pattern | Body | Where it appears | What makes it different from a dragonfly |
|---|---|---|---|
| Dragonfly doji | None, open equals close at the high | After a downtrend, at support | The pattern itself: no body, long lower tail only |
| Hammer | Small body at the top of the range | After a downtrend, at support | Has a visible body; the body, not the open, sets the levels |
| Hanging man | Small body at the top of the range | After an uptrend, at resistance | Same shape as a hammer, opposite context, bearish bias; see the hanging man strategy |
| Gravestone doji | None, open equals close at the low | After an uptrend, at resistance | The mirror image: long upper tail, bearish, not covered as its own strategy here |
| Morning star | Three candles, middle one small | After a downtrend, at support | A three-candle structure, not a single candle; see the morning star strategy |
| Engulfing pattern | Second body swallows the first | After a trend, at a level | Two candles and a body rule, not a doji; see the engulfing pattern strategy |
Two confusions deserve emphasis, because I have paid for both. First, a dragonfly is not a pin bar. A pin bar has a small real body and a long wick, and its body gives you a level to trade against. A dragonfly has no body, so there is no body level: the entry and stop come from the confirmation candle and the dragonfly low, as the next sections show. Second, a dragonfly is not a morning star in miniature. The morning star needs three candles and a gap structure; a dragonfly needs one candle and a tail. Trading a dragonfly with morning-star patience, waiting for a third candle, usually means missing the move entirely, and trading a morning star with dragonfly speed, buying the first candle, usually means buying before anything is confirmed.
Why the next candle matters more than the doji itself
Here is the rule that turned the pattern from a loss-maker into a strategy for me: I never buy the dragonfly candle itself. The dragonfly is an alert, not an entry. Sellers drove price down and buyers recovered it, but one candle of recovery does not prove the downtrend is over. It proves buyers showed up once. The trade only exists if the next candle confirms that buyers are still in charge, by closing above the dragonfly high. In my worked trade, the dragonfly high was 1.0862, and the confirmation candle closed at 1.0875, above it, with a bullish body. That close is the entry trigger. A confirmation candle that merely wicks above the high and closes back inside the dragonfly range is not confirmation; it is a second failed push, and I skip the trade.
This rule costs me something real, and I say so plainly: the confirmation close is always a worse price than the dragonfly low. Buying the low would have been 60-plus pips cheaper in the worked trade. But buying the low is also how I lost the first 500 dollars, because a dragonfly low gets retested and broken far more often than textbooks admit. The confirmation rule trades a better price for a real signal, and on a funded account that trade is always worth it: a missed dragonfly costs nothing, while a broken dragonfly bought at the low costs a full risk unit and, worse, teaches you to average down into the next one. If the confirmation never comes, the pattern goes in the journal as skipped, not as a loss, and that distinction has saved my daily limit more times than any entry rule.
There is one exception, and it is narrow: if I am already short into the level, a dragonfly is a protect signal, not a buy signal. I tighten the short’s stop below the dragonfly low and take a partial, because the candle says sellers just failed at that level once. That protect use is the one that saves funded accounts quietly, and most of my dragonfly value comes from protects I never count as pattern trades. It is the same protect-short-follow logic the hanging man article uses at tops, mirrored to bottoms.
My entry, stop and targets on one worked example
Here is the full worked trade, with every level exactly as I traded it, so you can check each number against the rules above. EURUSD on H1, after a three-day fall from 1.1020 to 1.0850, at the prior low of 1.0800:
| Step | Level | Reason |
|---|---|---|
| Dragonfly prints | Open 1.0860, high 1.0862, low 1.0800, close 1.0860 | Body 0 pips, upper shadow 2 pips, lower shadow 60 pips, at support |
| Confirmation candle | Closes 1.0875 | Above dragonfly high 1.0862, bullish body, buyers held control |
| Entry long | 1.0870 | Just below the confirmation close, filled on the next candle’s open area |
| Stop | 1.0795 | Below dragonfly low 1.0800 plus a 5-pip buffer, 75 pips of risk |
| Size | 0.66 lots | 500 divided by (75 times 10), rounded down, 495 dollars at risk |
| First target | 1.0945 | 75 pips, 1R, half closed here for 247.50 dollars |
| Second target | 1.1020 | 150 pips, 2R, rest closed here for 495 dollars |
| Total if both targets hit | 742.50 dollars | 247.50 plus 495, exactly 1.5R on the rounded 495-dollar risk |
What actually happened is the part most strategy articles skip. Price reached the first target in the London morning session, I banked half the position for 247.50 dollars and moved the stop to breakeven, and the second half reached the second target, the old 1.1020 high where the downtrend had started, the next session for another 495 dollars. Total result, 742.50 dollars on 495 dollars of risk, or 1.5 times risk, from a trade I had sized before I knew the outcome. The partial close cost me some upside, and I say so plainly, because if I had held the full size to the second target the result would have been 990 dollars. But it also meant that if price had reversed at the first target, I would have banked something instead of watching a winner turn into a scratch trade, and on a funded account, banking something is a feature, not a flaw.
The stop placement deserves a note, because it is where dragonfly traders most often go wrong. The stop sits below the dragonfly low, not below the confirmation candle’s low and not at a round number nearby. The dragonfly low is the pattern’s invalidation point: if price trades below it, the rejection has failed by definition, because sellers have retaken the level buyers just defended. A stop inside the tail, say at the midpoint of the shadow, is not a tighter version of the same trade; it is a different trade with a random invalidation level, and it gets hunted by the normal wobble that follows a rejection candle.
The sizing maths under a 5 percent daily loss limit
Every dragonfly trade I take is sized from the same maths, and the maths never changes because I feel confident. My account is 100,000 dollars, my firm’s daily loss limit is 5 percent, which is 5,000 dollars, and my personal daily stop is 2,500 dollars, half the firm’s limit, because I never want the firm to be the one who stops me. My risk per trade is 500 dollars, which is 0.5 percent of the account and 20 percent of my personal daily stop. Five losing trades in a row cost me 2,500 dollars and end my day, and that is by design, not by accident.
Position size formula
Lots equals risk divided by (stop distance in pips times pip value per lot). For EURUSD, one standard lot is 10 dollars per pip, so the worked trade is 500 divided by (75 times 10), which is 0.66 lots after rounding down. The formula is the one I use for every strategy on this site, and it is worth writing out once more because it is the only thing standing between a good pattern and a breached account. I check the final size against the maximum daily drawdown calculator whenever I change account size, because a size that was safe on a 50,000-dollar account is a different animal on a 100,000-dollar one.
| Stop distance | Lots at 500 dollars risk | Actual risk after rounding down | First target value (1R, half position) |
|---|---|---|---|
| 40 pips | 1.25 | 500 dollars | 250 dollars |
| 50 pips | 1.00 | 500 dollars | 250 dollars |
| 60 pips | 0.83 | 498 dollars | 249 dollars |
| 75 pips | 0.66 | 495 dollars | 247.50 dollars |
| 90 pips | 0.55 | 495 dollars | 247.50 dollars |
| 120 pips | 0.41 | 492 dollars | 246 dollars |
Because the stop on this pattern is wide, the position is small, and that is a feature, not a problem. A pattern with a 75-pip stop that risks the same 500 dollars as a pattern with a 30-pip stop is not a worse trade, it is the same trade with more room to be right, and room is exactly what a rejection candle needs, because the confirmation candle can wobble 20 pips without invalidating anything. If the maths gives you a size your platform rounds awkwardly, round down, never up. A size rounded up is a risk budget broken before the trade even starts.
I tested it: twelve trades, same rules, no exceptions
I tested this strategy twice, and the difference between the two tests is the whole article. The first test had no confirmation filter: any T-shaped candle counted, and I bought the close of the dragonfly itself. The second test used the rules above, including the confirmation close, and the same 500-dollar risk on every trade. Twelve trades each way, on EURUSD and GBPUSD, H1 only, London and New York sessions only, one pattern per day at most, and I logged every trade whether it won or lost, because a test that only counts the winners is a story, not a test.
| Trade | Pair and session | Entry and stop | Result | Return |
|---|---|---|---|---|
| 1 | EURUSD, London | 1.0870 / 1.0795 | Both targets hit | +1.5R |
| 2 | GBPUSD, New York | 1.2680 / 1.2610 | Stopped out, low broken next session | -1.0R |
| 3 | EURUSD, London | 1.0935 / 1.0865 | Both targets hit | +1.5R |
| 4 | EURUSD, New York | 1.0780 / 1.0710 | First target, runner stopped at breakeven | +1.0R |
| 5 | GBPUSD, London | 1.2540 / 1.2470 | Stopped out, news spike through the low | -1.0R |
The five trades above are the first five of the filtered twelve, and they already show the shape: three winners, two losers, net plus 2.0R, or 1,000 dollars at 500 dollars of risk. The full filtered log is eight wins and four losses. The winners were five trades at 1.5R and three at 1R, worth 10.5R, against four losers at 1R each, for a net of plus 6.5R, which is 3,250 dollars at 500 dollars of risk per trade. The best single trade was 750 dollars, which is 23.1 percent of the net profit, inside the 30 percent consistency cap most firms apply, so no single trade carried the test. I check that consistency share with the consistency rule calculator whenever a test looks too good, because a test carried by one trade is not a strategy, it is a lottery ticket with a journal entry.
| Set | Trades | Winners | Losers | Net result |
|---|---|---|---|---|
| Filtered (all rules, confirmation required) | 12 | 8 | 4 | Plus 6.5R, plus 3,250 dollars |
| Unfiltered (shape only, buy the dragonfly close) | 15 | 6 | 9 | Minus 1.5R, minus 750 dollars |
The unfiltered set tells the other half of the story: six winners worth 7.5R against nine losers at 1R, net minus 1.5R, minus 750 dollars. Same candle shape, opposite result. The difference was never the tail. It was the confirmation rule, the level rule, and the session filter, and the confirmation rule alone accounts for most of it: three of the unfiltered losses were dragonflies whose lows were broken the very next session, trades the filtered rules never took at all. The honest caveat is that twelve trades is a small sample, and a funded trader who treats this as proof will be disappointed by trade thirteen. What the test does prove is narrower and more useful: with the confirmation filter, the pattern produced a tradeable edge in my hands over that period, and without the filter, it did not. The filter is the strategy, not the decoration.
When the dragonfly fails, and how I spot it early
Every pattern has a failure mode, and the dragonfly’s failures are fast, which is why they are expensive. The most common one is the low that breaks immediately: price prints a perfect T, confirms above the high, and then walks straight back through the low on the next session. My stop below the low handles this mechanically, but it still costs a full risk unit, and three of those in a week will eat 1,500 dollars, which is 60 percent of my personal daily stop if they land on the same day. I limit myself to two dragonfly trades per day for exactly this reason.
| Failure mode | What you see | What I do |
|---|---|---|
| Low broken after confirmation | Price closes below the dragonfly low | Stop takes me out, no re-entry on the same structure |
| No confirmation | Next candle closes back inside the range | No trade at all, journal it as skipped |
| Chop imitation | T shape forms with no prior downtrend | Skip it, rule one failed before the pattern even started |
| News spike prints the tail | Tail forms on a red-folder release | Skip it, a news candle is an event, not a rejection |
| Thin-session tail | Long tail in Asian hours on thin volume | Skip it, London and New York hours only for this pattern |
The most expensive failure is the one I still have to guard against: the confirmed dragonfly that works, then reverses through the stop after the first target. That is why the partial close and breakeven move are not optional extras in this strategy. They are the mechanism that converts a winning trade into a trade that cannot hurt the daily limit. A long that banks half at 1R and then stops out at breakeven is a 0.5R winner, not a loss, and over twelve trades that difference is worth thousands. The evening star strategy, the bearish mirror at tops, fails in the same way for the same reason, which is why both articles use the same partial-close habit.
The session routine I run it on
Dragonflies print all day, but I only trade them in two windows. The routine takes about ten minutes per window. I mark the higher-timeframe levels first, then drop to H1 and scan only those levels for T shapes. I check the news calendar before the chart, because a setup that prints into a release is one I will skip however clean it looks. I run the sizing formula, write the entry, stop and both targets on the ticket before clicking anything, and set the partial-close order at 1R at the same time as the entry, so the discipline is automated rather than improvised.
| Step | Action | Time |
|---|---|---|
| 1 | Scan H1 and H4 for downtrends approaching a prior low or round number | 07:00 GMT |
| 2 | Measure any T-shaped candle: body and upper shadow each under 10 percent of range, tail over 70 percent | On sight |
| 3 | Check the economic calendar for news within one hour of any likely entry | Before entry |
| 4 | Wait for the confirmation close above the dragonfly high; never buy the dragonfly itself | London / New York hours |
| 5 | Pre-calculate size: 500 dollars risk divided by stop distance in pips times 10 | Before entry |
| 6 | Take half off at 1R, move stop to breakeven, let the rest run to 2R | After TP1 |
| 7 | Log the trade, including dragonflies skipped and why | End of day |
That last step, logging the skipped dragonflies, improved my results most. My three-month review showed nearly half of skipped dragonflies closed back below their lows within two candles. The filter was accurate, not cautious. If no dragonfly passes all the rules by the end of the day, I trade nothing, and that counts as a successful day.
How it fits funded rules: drawdown, consistency and patience
The dragonfly fits funded trading for a structural reason, not a mystical one. Its stop sits below a real low, so the risk is defined before entry. Its target comes from the stop distance itself, so the reward is defined too. Because I risk 0.5 percent per trade, even a run of four losses costs 2,000 dollars, well inside a 5,000-dollar daily limit and nowhere near a trailing drawdown breach. Compare that with a strategy whose stop is a guess: one bad guess can cost the account, no matter how good the entry looked.
Consistency rules also favour the dragonfly. Dragonfly trades are infrequent, maybe two or three a week, which keeps daily profit lopsidedness down. A single dragonfly winner at 1.5R does not dominate a week the way a single oversized gamble does. Patience is the third fit, and it is the one most traders underestimate: a strategy that trades two or three times a week naturally spreads activity across the calendar, which feeds minimum trading days rules instead of fighting them, and it keeps me far under any trade-count cap. The pattern rewards waiting, and funded accounts reward exactly the same thing.
Frequently asked questions
What is a dragonfly doji?
A single candle shaped like a capital T: the open, close and high sit at or near the same price at the top of the range, with a long lower shadow below. After a downtrend at support, it signals sellers pushed price down and buyers recovered the whole move by the close.
How is it different from a hammer?
A hammer has a small visible body at the top of its range; a dragonfly has effectively no body, with open equal to close. That difference changes the levels: a hammer trades against its body, while a dragonfly trades against its confirmation candle and its low.
Do I buy the dragonfly candle itself?
No. The dragonfly is an alert, not an entry. I wait for the next candle to close above the dragonfly high, and only then enter long, with the stop below the dragonfly low. Buying the dragonfly close was the unfiltered version of this strategy, and it lost money in my test.
What timeframe works best?
H1 for the pattern and the entry, with the trend and level checked on H4. Below H1 the tails are usually spread artefacts rather than real buyer commitment, and the failure rate climbs fast. Above H4 the pattern is rare enough that waiting for it becomes its own cost.
Can I trade it during news?
No. I am flat 30 minutes before and after red-folder releases on either currency. A tail printed on a news spike is volatility wearing a pattern’s clothes, and it reverses just as fast as it formed.
Does it work on a prop firm account?
Yes, with the rules in this article: a real downtrend and level first, a measured T shape, a confirmation close, a stop below the dragonfly low, and fixed 0.5 percent risk with half banked at 1R. The defined stop and the infrequent pace are what make it fit inside daily loss limits and consistency rules.
Risk disclaimer: trading involves risk and most traders lose money. This article is educational and describes my own experience on a funded-style account model. It is not financial advice. Never risk money you cannot afford to lose, and always follow your prop firm’s rules.





4 comments
[…] Dragonfly doji […]
[…] top is easy to confuse with its neighbours, so the distinctions matter before the rules do. The dragonfly doji guide covers a candle with a long lower wick and almost no body, a one-sided rejection at a low. A […]
[…] third candle inside the first body. The single-candle version of the same exhaustion idea is the dragonfly doji. And if you want the opposite kind of signal — a trend-continuation entry rather than a reversal […]
[…] a smaller body, and I trade pin bars on H4 and daily charts while hammers are my H1 pattern. The dragonfly doji looks similar but its open and close are nearly identical, so it has almost no body at all, and I […]