A stock has been bleeding for eight sessions. On the ninth it opens, gets hammered again, drops hard through the morning, and then something odd happens. Buyers soak up everything. By the close the price is back almost exactly where it opened, right at the top of the day’s range. That candle, a tiny body perched on a very long tail, is what candlestick traders call a takuri line.
Most people would just call it a hammer. They’re not wrong, but they’re not quite right either. The takuri line is a stricter, more extreme version, and the two main references that define it don’t even agree on how strict. This piece goes through both definitions, the test numbers behind the pattern, and three worked examples so you can see where the rules bite.
In this article
- What a takuri line looks like
- What the takuri line test numbers say
- Worked example 1: does this candle qualify?
- Worked example 2: a hammer that isn’t a takuri to the scanner
- Worked example 3: turning the candle into a trade plan
- Takuri line vs hammer vs dragonfly doji
- Mistakes we see with this pattern
- Quick recap
What a takuri line looks like
Strip it down and you need three things:
- A downtrend going in. The candle only means something if the stock has been falling.
- A small real body at the top of the range. Open and close are close together, and both sit near the day’s high.
- A very long lower shadow, with little or no upper shadow. The tail is the whole story. Sellers pushed the price far down during the session and could not keep it there.
Thomas Bulkowski, who tested 103 candle types for his Encyclopedia of Candlestick Charts, gives the configuration on his takuri line page as a small-bodied candle with a lower shadow at least three times the height of the body and little or no upper shadow, appearing after a downward price trend. He adds that the body can be any colour. A green takuri and a red takuri are read the same way.
Compare that with his description of the hammer: a long lower shadow at least two or three times the height of the body, with little or no upper shadow, in a downtrend. So every takuri is a hammer. Not every hammer is a takuri. The takuri simply insists on the longer tail.
The software version is stricter still
If you run a scanner or a backtest, there’s a good chance the pattern code underneath comes from TA-Lib, the open-source technical analysis library. Its function list names the function CDLTAKURI and labels it “Takuri (Dragonfly Doji with very long lower shadow)”. That word doji matters. TA-Lib doesn’t just want a small body. It wants a body so small it barely exists.
The CDLTAKURI source code checks for a doji body, no or a very short upper shadow, and a very long lower shadow. What those words mean is set in the library’s default candle settings:
- Doji body: the real body is no bigger than 10% of the average high-low range of the previous 10 candles.
- Very short upper shadow: shorter than 10% of that same 10-candle average range.
- Very long lower shadow: longer than two times the candle’s own real body.
Notice the twist. TA-Lib’s shadow test is only “more than twice the body”, which sounds looser than Bulkowski’s three times. In practice it’s the body test that does the work. Once the body has to be under a tenth of a normal day’s range, almost any visible tail clears the twice-the-body bar. Also notice what TA-Lib doesn’t check: the downtrend. The code looks only at the shape of the candle, so a scanner using it will flag takuri shapes in uptrends and sideways markets too. Filtering for trend is your job.
Every takuri is a hammer, but not every hammer is a takuri. And the scanner you rely on may be using a stricter definition than the book you learned from.
What the takuri line test numbers say
Bulkowski’s opening line on the pattern is “Finally, a single candle line that works!” Here’s what he reports, all ranks out of 103 candle patterns where 1 is best:
- Reversal rate: it acts as a bullish reversal 66% of the time. That is the share of tested patterns where the prior downtrend reversed, which puts it 18th of 103.
- Frequency rank: 28, so it turns up often enough to be practical.
- Overall performance rank: 47, roughly mid-table. In his words, price may reverse often, but it does not trend far afterwards.
- Price target: projecting the candle’s height in the breakout direction, price hit that target 82% of the time in the best case, which was an upward breakout in a bull market.
- Best 10-day move: a fall of 4.45%, after a downward breakout in a bear market.
Put those next to his hammer figures. The hammer reverses 60% of the time (reversal rank 26), has a frequency rank of 36 and an overall performance rank of 65. Its best case hits the height target 88% of the time, and its best 10-day move is a 4.12% drop after a downward breakout in a bear market.
So the longer tail buys you something. The takuri reverses more often than the plain hammer and ranks better overall. But 66% is still one failure in every three. And the best 10-day move for both patterns came on the downside, which tells you what happens when the support from that long tail gives way. One more caveat: the page doesn’t say the testing was done on Indian stocks, so treat these as a rough guide rather than a fact about NSE charts.
Bulkowski also passes on three practical tips from the book. Look for the takuri as part of a pullback inside a larger uptrend. Wait for price to close higher the next day before treating it as confirmed. And takuri lines within a third of the yearly high tend to act as reversals most often.
Worked example 1: does this candle qualify?
Take a hypothetical mid-cap that has slipped for two weeks. One day it prints:
| Open | High | Low | Close |
|---|---|---|---|
| ₹612.40 | ₹614.00 | ₹584.00 | ₹613.20 |
Work out the pieces:
- Real body = 613.20 − 612.40 = ₹0.80 (a green candle).
- Upper shadow = 614.00 − 613.20 = ₹0.80.
- Lower shadow = 612.40 − 584.00 = ₹28.40.
- Total range = 614.00 − 584.00 = ₹30.00.
Bulkowski’s test: the lower shadow is 28.40 ÷ 0.80 = 35.5 times the body. Far more than three. The upper shadow is ₹0.80 on a ₹30 range, so “little” is fair. It came after a downtrend. It qualifies.
TA-Lib’s test needs one more input, the average high-low range of the previous 10 sessions. Say it was ₹14.00. Then the doji limit is 10% of that, ₹1.40. The body of ₹0.80 is under it. The upper shadow of ₹0.80 is under ₹1.40, so it counts as very short. The lower shadow of ₹28.40 is more than twice the body (₹1.60). Flagged.
Both definitions agree. This is the textbook case.
Worked example 2: a hammer that isn’t a takuri to the scanner
Now a different hypothetical stock:
| Open | High | Low | Close |
|---|---|---|---|
| ₹1,040 | ₹1,049 | ₹1,010 | ₹1,048 |
- Real body = 1,048 − 1,040 = ₹8.
- Upper shadow = 1,049 − 1,048 = ₹1.
- Lower shadow = 1,040 − 1,010 = ₹30.
The lower shadow is 30 ÷ 8 = 3.75 times the body, so by Bulkowski’s rule this is a takuri line. Suppose the 10-session average range is ₹25. TA-Lib’s doji limit is then ₹2.50, and an ₹8 body is more than three times that. CDLTAKURI returns nothing.
Same candle, two answers. That’s not a bug. It’s two people drawing the line in different places. If you read a backtest that says “takuri lines on Nifty stocks did X”, the first question to ask is which definition the code used. A doji-bodied takuri is a much narrower filter than a small-bodied one, so the two backtests are effectively testing different patterns.
The long tail shows sellers failed for one day. Only the next close shows whether buyers can actually take charge.
Worked example 3: turning the candle into a trade plan
Go back to the first stock. We have a takuri at the end of a decline. What now?
Confirmation. Following Bulkowski’s tip, we don’t buy on the takuri day itself. We wait for a higher close the next day. In his chart examples, he treats the breakout as upward once price closes above the top of the candle, here the high of ₹614.00. Say the next session closes at ₹617.50. That’s our entry.
Stop loss. The whole idea rests on that ₹584 low holding. If price trades below the bottom of the tail, the signal has failed. Put the stop just under it, at ₹583.
Target. Use Bulkowski’s height projection: candle height ₹30.00 added to the top of ₹614.00 gives ₹644.00.
Now do the arithmetic most people skip:
- Risk per share = 617.50 − 583.00 = ₹34.50.
- Reward per share to target = 644.00 − 617.50 = ₹26.50.
- Reward-to-risk = 26.50 ÷ 34.50 ≈ 0.77.
That’s the uncomfortable part. A very long tail makes a very wide stop. The measured target only gets you back roughly one candle height, and by the time you have confirmation you’ve already given up some of it. You’d be risking more than you stand to make on the first target. This lines up with Bulkowski’s finding that price reverses often but doesn’t travel far.
Position size. Say the trading account is ₹5,00,000 and you cap the loss on any one trade at 1%, or ₹5,000. Then 5,000 ÷ 34.50 = 144.9, so 144 shares. At ₹617.50 that’s a position of ₹88,920, and if the stop is hit you lose 144 × 34.50 = ₹4,968, inside the cap.
Given that 0.77 ratio, a trader has three honest choices. Trail the stop up once the target is hit and let the rest run, hoping for one of the larger moves. Take the trade only when a bigger-picture level, such as an old breakout point or a rising moving average, sits near the tail’s low. Or skip it. Skipping is a valid trade.
Location. Finally, the yearly-high tip. Suppose the stock’s 52-week range is ₹480 to ₹700. A third of that ₹220 range is about ₹73.33, so the top third starts near ₹626.67. Our takuri closed at ₹613.20, below that zone. So this one doesn’t get the extra boost Bulkowski describes. It might still work, but it’s not the best version of the setup.
Takuri line vs hammer vs dragonfly doji
| Hammer | Takuri line (Bulkowski) | TA-Lib CDLTAKURI | |
|---|---|---|---|
| Body | Small | Small, any colour | Doji: ≤10% of 10-candle average range |
| Lower shadow | At least 2–3× body | At least 3× body | More than 2× body |
| Upper shadow | Little or none | Little or none | Under 10% of 10-candle average range |
| Downtrend check | Yes | Yes | No, add it yourself |
TA-Lib also has a separate CDLDRAGONFLYDOJI function. Its label for the takuri, “Dragonfly Doji with very long lower shadow”, tells you how the library sees it: the long-tailed end of the dragonfly family, with the code’s own comment asking for the open and close at the high of the day. If you want the mirror image of that shape, with the tail pointing up, our piece on the gravestone doji covers it.
Mistakes we see with this pattern
- Calling every long tail a takuri. Measure it. If the lower shadow isn’t three times the body, it’s an ordinary hammer at best.
- Ignoring the trend. The same shape after a rally is a hanging man, a warning sign, not a buy signal. Scanners built on TA-Lib won’t make that distinction for you.
- Buying the close of the takuri day. The confirmation tip exists because one in three of these fails.
- Using a tight stop to make the numbers look better. A stop placed halfway up the tail sits inside the very range where sellers were already beaten back once. It gets hit on noise.
- Trusting a backtest without checking the definition. As example 2 showed, a doji-bodied takuri and a small-bodied one are different filters.
If you’re newer to candles, start with our hammer candlestick pattern guide, since the takuri is just its stricter cousin, and the wider candlestick patterns for beginners list for the rest of the family.
Quick recap
- A takuri line is a single candle after a downtrend: small body at the top, lower shadow at least three times the body (Bulkowski), little or no upper shadow.
- TA-Lib’s CDLTAKURI demands a doji body and a very short upper shadow, and doesn’t check the trend.
- In Bulkowski’s tests it reversed the downtrend 66% of the time versus 60% for the hammer, but the follow-through was only mid-table.
- Wait for a higher close, put the stop under the tail, and check whether the reward justifies a wide stop before you take it.
Disclaimer: This article is for education only and is not investment advice or a recommendation to buy or sell any security. The examples use hypothetical prices. Candlestick patterns fail often and past test results don’t guarantee future outcomes. Do your own research or consult a SEBI-registered investment adviser before trading.





Leave a Reply
Join the conversation
Commenting is open to registered readers. Sign up with just your email — no password needed, and it takes under a minute.
Sign up to comment Log in