A closing marubozu is a long candle with no shadow at the end where it closed: a green one that closes at its high, a red one that closes at its low. A small shadow at the opening end is allowed.

Two things decide whether your count matches anyone else's, and the guides settle neither. Is it one end, or both? And how short does "no shadow" have to be?

What is a closing marubozu?

It is a single candle. The closing end has no shadow, or almost none, and the body is long. TA-Lib's CDLCLOSINGMARUBOZU page puts the formula in two parts: "(1) long real body: real body > the BodyLong average; AND (2) very short shadow at the closing end: if white (close>=open) upper shadow < the ShadowVeryShort average [close at/near high]; if black (close<open) lower shadow < the ShadowVeryShort average [close at/near low]".

The name comes from Japanese. Alchemy Markets puts it plainly: "marubozu is the Japanese term for bald hill or closed-cropped (shaved) head". The closing version is bald at one end only.

The same TA-Lib page says what the candle is: "A strong directional bar, not a defined reversal/continuation signal".

Is the closing marubozu built into ChartLabs?

Yes, but you have to switch it on. Open the Indicators menu and, in the Standard list, add Candlestick Patterns from the Patterns heading. It draws on the price pane and holds 61 pattern switches. Closing Marubozu is in the one-candle group and is off when you add the indicator, so turn it on in the indicator's settings. Marubozu, with no shadow at either end, is a separate switch and is off by default too.

The test is TA-Lib's: a long body and almost no shadow at the closing end, both measured against the ten candles before it. Those thresholds are TA-Lib's defaults, and you cannot change them. A bullish one is marked CM below the low, a bearish one above the high. There is no scanner and no price alerts.

Is a closing marubozu the same as a marubozu?

No. There are three shapes in the family, and the difference is which end the shadow is missing from.

Name	Shadow at the open end	Shadow at the close end
Full marubozu	none	none
Closing marubozu	a small one allowed	none
Opening marubozu	none	a small one allowed

TA-Lib implements the first two as separate functions. CDLMARUBOZU is described as a "Single candle with a long real body and no/very-short upper and lower shadows, so open and close sit at the range extremes". Its function index lists 61 CDL functions, including CDLCLOSINGMARUBOZU, CDLMARUBOZU and CDLBELTHOLD, and no CDLOPENINGMARUBOZU. The opening variant has a name in the literature and no function in that library.

Snappchart gives the two partial versions in one line each: "An opening marubozu has no wick at the open end but a small one at the close end... A closing marubozu has no wick at the close end but a small one at the open end". LiteFinance says the same thing from the other direction: "Marubozu close candle has no shadow on the closing price side and has a small wick on the opening price side".

The belt hold that sits next to these in the index is a separate pattern with its own definition. LiteFinance names it in passing: "This pattern is similar to Japanese candlesticks such as bullish or bearish counterattack and belt hold. They all have a large real body that has short or no shadows".

How do you tell a bullish from a bearish closing marubozu?

The colour of the body, not the missing end. A green candle closes at its high, so the missing shadow is the upper one. A red candle closes at its low, so the missing shadow is the lower one.

Two guides still get it the wrong way round:

HDFC Sky is correct in its numbered section, "It shows no shadow on the closing price side. It has a brief wick higher or below the opening prices", and then its own FAQ answers the opposite: "A bullish Marubozu close is when a candlestick has no lower shadow, and the closing price is at the trading session's highest point". A green bar with no lower shadow and a close at the high has no shadow at either end, which is the full marubozu.
Get Together Finance has it right in the body, "if a candle has no wick at the top but a small one at the bottom, it's a Bullish Close Marubozu", and reversed in its FAQ: "It occurs when the opening price of a stock is below the closing price, with the equal of closing price to the high. This pattern shows a robust bearish sentiment, recommending sellers control the complete session". That describes a green bar closing at its high and labels it bearish.

The same page can be right in its body and wrong in its FAQ, so read both before you copy a rule out of one.

How short does the shadow have to be to count?

TA-Lib publishes a figure, the guides do not, and the answer changes the count by about four times.

The library's candle settings page gives both of the thresholds its formula names. BodyLong is the mean real body of the previous 10 candles, taken at a factor of 1.0. ShadowVeryShort is the mean high-to-low range of the previous 10 candles at a factor of 0.1, so a tenth of that average range. None of the guides puts a number on "no shadow".

So pick one and write it down. On 16,510 EURUSD 15-minute bars from our own data, with the closing-end shadow measured as high minus close on a green bar and close minus low on a red one:

Closing-end shadow	Bars (closing end only)	Bars (both ends)
Exactly zero	1,044	93
Up to 1% of the bar's range	1,106	100
Up to 5%	2,532	436
Up to 10%	4,302	1,205

From the strictest reading to a 10% shadow is 1,044 to 4,302, a little over four times as many bars, from one word the guides leave open. The percentages in the table are ours.

Run the library's own defaults and you get 2,478 bars on the same EURUSD 15-minute candles: a long body above the mean body of the previous 10 bars, plus a closing-end shadow under 10% of the 10-bar mean range. Those are its published BodyLong and ShadowVeryShort settings, so this is the library's rule on our data.

Why does it matter whether you read one end or both?

Because the both-ends reading throws away about nine in ten of the bars the term names. On the same 16,510 EURUSD 15-minute bars, the exact closing-end reading gives 1,044 and the both-ends reading gives 93. That is 11.2 times fewer, and 91% of the closing marubozus gone. On hourly bars it is 109 against 6.

One guide prints the both-ends rule under this name, and contradicts itself doing so. Patterns Wizard's summary line is right: "It's a long candle close at it's high (bullish) or low (bearish)". Its definition section then says "The bullish closing marubozu candlestick pattern forms when the open price is the low price and the close price is the high price of the day", with the bearish bullet mirroring it, and its identification section adds "It signifies that the body of the candle does not have upper or lower shadow (or a small shadow visible on the trading charts)". Those are three statements of the full marubozu under the closing marubozu's name.

HDFC Sky's FAQ answer, quoted above, lands in the same place by naming the open end as well as the close.

The usual failure on this term is smaller and harder to catch: a correct main text with a wrong FAQ, a wrong end, or a wrong label.

What does a closing marubozu mean in an uptrend?

One bar does not tell you, and two of the sources say so.

Huntraders defines it at one end only ("Bullish Marubozu Closing has only lower shadow", "The closing price is the daily maximum. Thus, there is no upper shadow"), lists "Reliability: Low" as a property, and writes: "Bullish Marubozu Closing is not able to forecast market movements. Depending on the previous day's candle, it can be both a bullish confirmation and a bearish reversal pattern... It is necessary to observe previous candles to determine the true role of the Bullish Marubozu Closing".

Set that next to TA-Lib's "A strong directional bar, not a defined reversal/continuation signal". A green closing marubozu in an uptrend says buyers held the bar into the close. It says nothing about the next bar.

Three guides go further than their own evidence:

Elearnmarkets turns it into a forecast: "Therefore, it is expected that when the trading session reopens, the price will continue moving in the direction of the close".
Snappchart ranks the two partials: "The closing version tends to read stronger, since finishing at the extreme of the bar is a better tell than starting there", stated as fact.
LiteFinance describes the meaning both ways in one sentence, "often hinting at trend continuation or preparing traders for a possible trend reversal", without saying how to tell which.

Huntraders and the library both stop short of a forecast.

How often does a closing marubozu print?

Often enough on 15-minute bars to test, and barely at all on daily bars.

We ran both readings over our own cleaned data on two instruments: EURUSD, 246,778 one-minute bars, and GER40, the euro-quoted German 40 index, 225,951 one-minute bars. Both cover 1 January to 31 August 2026, aggregated from 1-minute candles to 15-minute, hourly, four-hour and daily bars. We source from Dukascopy, tick-level where it is available, clean it to remove anomalies, and convert to 1-minute candles as our lowest timeframe. Every count below runs after a 10-bar warm-up for the body average, so all of them sit on the same bars.

Cell	Bars	Closing end, exactly zero shadow	Both ends
EURUSD M15	16,510	1,044	93
EURUSD H1	4,120	109	6
EURUSD H4	1,058	10	1
EURUSD D1	198	1	0
GER40 M15	15,171	423	29
GER40 H1	3,887	51	8
GER40 H4	1,031	10	1
GER40 D1	197	1	0

One occurrence on 198 EURUSD daily bars is one occurrence, across eight months. It is not a rate and it does not generalise. These are occurrence counts, and none of them says what price did after any of those bars.

The body test and the shadow test bite independently, and those counts apply the shadow test only: of the 1,044 EURUSD 15-minute bars that close with no shadow at all, 615 also have a long body, and those same 615 are all that survives of the 6,830 long-bodied bars once the shadow test runs. Each condition removes most of what the other one keeps, which is why the library asks for both.

GER40 runs on nearly the same clock as the pair, pausing for about an hour each evening around 21:00 UTC, and its counts come out the same shape as the forex ones, smaller throughout.

How do you test a closing marubozu rule step by step?

Write the shadow threshold down first, in words and in numbers. Exactly zero, or a percentage of the bar's range, or a percentage of a recent average range.
Write the body rule down too. "Long" needs a definition. The library's BodyLong default is a real body above the mean real body of the previous 10 candles, and that is the one these counts use.
Decide which end you are measuring, and say so out loud. One end is the closing marubozu. Both ends is the full marubozu, and it is a different population.
Step bar by bar through a date range you picked before seeing any results, marking candidates as they close rather than scrolling back for them.
Record the threshold on every entry, not once at the top of the sheet. Two sessions a week apart will not use the same threshold unless it is written on each trade.
Do not mix timeframes in one count. On our bars the same rule gives 1,044 hits on 15-minute candles and 1 on daily candles.

What should you record for each one?

The shadow threshold, the end you measured, and the body rule. Without those three your count is not comparable with anyone else's, including the library's.

Then the usual: symbol, timeframe, date range, what the previous candles were doing, and what you did next. Huntraders' point about the previous candle is the context this pattern needs.

Marking one yourself in ChartLabs

ChartLabs gives you bar-by-bar replay on cleaned data, so a threshold rule can be watched hitting and missing in the order the market produced it. If your body and shadow test differs from TA-Lib's, the Pine Script editor will mark it: "Example: Mark bars by a rule" on the Pine Scripts tab opens a copy you can edit, and a script that encodes your own test can mark the bars it finds with a plotted shape or a label. There is no Pine backtesting. Written into a script, a threshold also stops drifting between sessions, and the trades CSV export carries a note per trade, so the version you used stays attached to what you traded.

Pick a threshold and test it

Decide what counts as no shadow, write it down, and step through a few months of 15-minute bars. The count you get is only meaningful next to that number.

Start a free trial and replay the bars yourself.

Closing Marubozu Candle: Which End Has No Shadow