10 min read The Chartix Editorial Team

Most articles on candlestick patterns for intraday trading are the daily-chart article with “5-minute” pasted into the headline. They list the same hammers and dojis and stars as if the timeframe changed nothing. It changes a great deal, and the honest version of this topic is not a list of the best patterns but a sorting: which pattern definitions still describe a real event when the candle is five minutes long, and which are left describing nothing.

This is the mirror of the argument in do candlestick patterns work on crypto. That post is about what happens when a market never closes. This one is about what happens when you compress the candle until it no longer contains a session at all.

What a 5-minute candle is, and what it is not

A daily candle on a stock summarises a session: an open that follows a night of accumulated news and an auction to absorb it, hours of continuous trading, and a close that people position for. Every part of that structure carries information, and several classic patterns are built on it.

A 5-minute candle is a five-minute slice of the same order flow. Nothing opens at 10:35 and nothing closes at 10:40. The “open” of that candle is the first print after the clock ticked over, which on a liquid continuous chart sits right on top of the previous close, and the “close” is whatever traded last before the next tick over. The four prices are recorded the same way and drawn the same way, but two of them, the open and the close, have lost most of the meaning they carry on a session chart. Nobody decided anything at the boundary. The boundary is a clock.

That single fact sorts the patterns. Anything whose definition is about the range and the close relative to the previous candle’s body still describes something. Anything whose definition leans on the open being a separate event from the previous close describes something that almost never happens intraday.

The patterns that carry over

These are the patterns whose definitions are purely about bodies overlapping and where the close landed. Compression shrinks the event, but it does not change what the shape means.

The engulfing patterns. Bulkowski’s identification for the bullish engulfing is a white candle whose open is below the prior close and whose close is above the prior open, body covering body, shadows ignored. On a continuous chart the “open below the prior close” part is nearly automatic, so the whole test collapses to one question: did this candle’s body cover the last one’s? That question is as meaningful in five minutes as in a day. It says the buying in this slice undid the selling in the last slice, which is a checkable statement about who was in control.

A bullish engulfing drawn from the pattern's real open, high, low and close proportions Down candle Engulfs it
Bullish engulfing, 2 candles, after a decline. Drawn to the real proportions the formation rules describe. Full bullish engulfing guide
  • Needs: a sequence of lower swing lows before it, not merely three red candles (more on what counts as a trend below).
  • Invalidated by: a close below the engulfing candle’s low. The bearish engulfing mirrors this, invalidated by a close above its high.

The marubozu. A marubozu is a candle with a full body and no wicks, or only negligible ones: it opened at one extreme of its range and closed at the other. Bulkowski describes it in the same terms, a tall candle with no shadows. The definition mentions no gap, no prior open, nothing but the candle’s own range, so it survives compression cleanly. On a 5-minute chart it is one of the few shapes that really is a statement about the last five minutes: one side took every print. Whether that matters depends entirely on where it happened, which is the section on context below.

A marubozu drawn from the pattern's real open, high, low and close proportions Marubozu
Marubozu, one candle, in any trend. Drawn to the real proportions the formation rules describe. Full marubozu guide
  • Needs: a location. A marubozu in the middle of a range is a burst; a marubozu that closes through a level the higher timeframe cares about is a different observation.
  • Invalidated by: a close back through the marubozu’s opening extreme, which fully retraces a session nobody contested and says the conviction was not shared.

One caution the intraday crowd tends to skip: Bulkowski’s own daily-chart figures have the black marubozu acting as a continuation 53% of the time and the white marubozu 56%, both of which he describes as near random. Compressing the timeframe does not improve on that. What the marubozu gives you intraday is a clean, unambiguous record of one-sided flow, not a forecast of what follows.

The three-candle sequences. Three white soldiers are three tall white candles, each opening inside the prior body and closing near its own high with higher closes; three black crows are the inverse. The “opens inside the prior body” condition is automatic on a continuous chart, exactly as the crypto post notes, so the entire content of the pattern lives in three consecutive closes near the extremes. That is fifteen minutes of one-directional flow with no meaningful pullback, which is a real thing to observe.

  • Needs: the prior move against the sequence. Three soldiers after a decline of several swings are a reversal candidate; three soldiers in the middle of a rise are just the rise.
  • Invalidated by: a close below the first soldier’s open (or above the first crow’s open for the bearish version).

The patterns that lose their definition

Anything that requires a gap. Bulkowski’s identification for dark cloud cover is a tall white candle followed by a black candle whose opening price is above the prior high, closing below the midpoint of the white body. On a 5-minute chart the open of one candle is the last print of the previous one. An open above the prior candle’s high requires price to jump between two consecutive prints, which in continuous trading is rare outside a halt or a news release. The one place it happens routinely is the first candle of the cash session, where the open is an auction print rather than a continuation of anything. So dark cloud cover, in its defined form, has almost no intraday instances except across the open.

The gapping variants of the morning star and evening star go the same way. The non-gapping versions still form, and they still describe a stall, but you are looking at the weaker instance and should weigh it as one. The crypto post makes the identical point about a market with no session boundary at all; intraday, you have a session boundary, but only one per day, and it is at the wrong end of the chart to help.

Exact dojis. A doji needs the open and close to be effectively equal. On a daily candle that is a full session of contest that ended where it began, a genuine standoff. On a 5-minute candle in a quiet stretch, it frequently means almost nothing traded in those five minutes, and the open and close match because they are nearly the same print. That is a liquidity artefact wearing a pattern’s clothes. The gravestone and dragonfly variants are somewhat better off because their wick is a real excursion, but the “closed where it opened” part still carries less weight than it does on a session.

Tweezers. The tweezer top and tweezer bottom are two adjacent candles with the same or nearly the same extreme. Bulkowski already finds the tweezers top near random on daily data, breaking out upward 56% of the time against its bearish theory. Intraday the matching problem gets worse, not better: two consecutive 5-minute candles printing the same high within a tick is often the same resting order being touched twice, or one participant working a level, rather than the market twice rejecting a price. The pattern is not wrong intraday. It is simply very easy to see where none exists, and two matching extremes on a fast chart should be read as a location marker, not evidence on their own.

The first and last half hour

This is where the intraday chart actively misleads, and the mechanism is worth stating plainly rather than as a rule about the clock.

The US cash session runs 9:30 a.m. to 4:00 p.m. ET, with an opening auction at 9:30 and a closing auction at 4:00. Nasdaq’s opening cross is a single-price auction: orders can be entered from 4:00 a.m., imbalance information is disseminated from 9:25, and at 9:30 the accumulated buy and sell interest is matched at one clearing price. NYSE’s pre-opening queues orders from 6:30 a.m. against the same 9:30 open.

The consequence for a 5-minute chart is that the first candle of the day opens at an auction print that reflects hours of queued orders and overnight news, and the minutes that follow are continuous trading discovering where the market clears relative to that print. A long wick on the 9:30 or 9:35 candle is usually that discovery process. Price overshot the auction level and came back, or the auction level was wrong and price left it. Neither is a rejection in the sense the hammer or shooting star describes, where a price was tested during normal two-sided trading and refused. The shape is identical; the process that made it is not.

The close has its own version. Nasdaq disseminates closing imbalance information from 3:50 p.m. ahead of the 4:00 cross, and orders that specifically request execution at the closing price are being matched at that single point. Candles in that window carry flow that is positioning for one print rather than expressing a view about the level, and the same shape that would mean something at 1:00 p.m. means less there.

None of this is a rule to avoid the first or last half hour. It is a description of why the patterns’ usual interpretation is weakest exactly when the shapes are largest, which is when the temptation to read them is strongest.

What counts as a trend on a 5-minute chart

Every entry above says “needs a prior trend,” and this is the part that fails most often intraday. On a daily chart, three red candles is three sessions of selling, which is three trading days of accumulated opinion. On a 5-minute chart, three red candles is fifteen minutes. That is not a trend. It is a pullback, or a lunchtime drift, or one large order being worked.

A prior trend on a fast chart means a sequence of swings: a lower high followed by a lower low, followed by another of each, so that the structure is visible when you zoom out. The candles are the pixels; the swings are the picture. A bullish engulfing after four consecutive lower lows and lower highs is forming after a decline. The same candle after three red bars in a range is the range doing what ranges do, and reversal candlestick patterns that have nothing to reverse are just candles.

The second half of the context requirement is that the higher timeframe decides whether the location matters at all. A 5-minute hammer is far more interesting when its low is sitting on a level the hourly or daily chart already marked, because then the intraday shape is describing how the market behaved at a price it had reason to defend. Support and resistance levels drawn on the higher timeframe are what turn a 5-minute shape from “a candle” into “a candle at a place.” Without that, most intraday patterns are describing the local behaviour of a handful of orders.

An illustrative example, with made-up prices. A stock has drifted from 50.80 to 50.20 over ninety minutes in a series of lower swings, and the daily chart shows a prior swing low at 50.15. The 5-minute candle that touches 50.12 and closes at 50.34, with a body that fully covers the previous red body, is a bullish engulfing forming after a genuine decline at a level the higher timeframe cares about. Its invalidation is a 5-minute close below 50.12. The identical candle at 50.50, in the middle of the drift, with no level under it, has the same shape and much less to say. Confirmation works the same way it does on any timeframe, and the confirmation candle post covers what the next candle must do; on a 5-minute chart that next candle arrives in five minutes, which is faster but not different.

Where this leaves the list

The sorting, in one place:

  • Carry over cleanly: engulfing patterns, marubozu, three white soldiers and three black crows. Their definitions are about bodies, ranges and closes, all of which a 5-minute candle records honestly.
  • Carry over with a heavier context requirement: hammers, shooting stars, spinning tops and the wicked dojis. The shapes mean what they always meant, provided the wick formed in two-sided trading and not in the auction windows, and provided there was a real swing structure before them.
  • Mostly stop existing: dark cloud cover, the gapping star variants, exact dojis on quiet stretches, and tweezers as standalone evidence.

One last point about the numbers. The reversal rates that get quoted for these patterns, Bulkowski’s 63% for the bullish engulfing included, describe the pattern as he defined and counted it. None of that transfers to a chart where the candle is a five-minute slice, and nothing in this post should be read as if it did. What transfers is the discipline, which is the same on every timeframe: name what the pattern requires before it means anything, and name the price that proves the read wrong.

Chartix reads a 5-minute screenshot exactly as it reads a daily one. It identifies the shape, states the context the pattern needs, and states the invalidation level in the same words, because the definition is the definition regardless of the clock. Whether a 5-minute chart is a sensible place to be reading candles at all is a question about you and your time, not about the patterns, and it is not one this post answers.

If you are still getting comfortable with what the four prices in a candle record, how to read a candlestick is the place to start, and how to read a stock chart covers why timeframe changes the picture. The full set of patterns with their context requirements and invalidation levels drawn out is in the candlestick pattern library.

Sources

Every figure above was checked against these before publishing.