What is a doji candlestick?

Every candlestick is a compression. Thousands of trades across a period are boiled down to four numbers: the open, the high, the low and the close. The real body, the distance from open to close, is the summary verdict; it measures how far one side managed to drag price away from where the period began. The wicks, sometimes called shadows, are the record of failed attempts: prices that traded during the period but could not hold into the close. A long green body says buyers won the period and kept their ground; a long lower wick says sellers pushed price down and lost every rupee of it back.

A doji is the special case where the verdict is a tie. Price left the open, in one direction or both, sometimes violently, and by the close it had been dragged all the way back. Sellers spent real money driving it down and kept nothing. Buyers spent real money driving it up and kept nothing. The wicks can be long, the traded volume can be heavy, the intraperiod swings can be wide, and the net result of all of it is zero. That combination, high effort and no displacement, is the entire information content of the candle. A trending bar says one side won the argument; a doji says the argument ended precisely where it started.

Notice carefully what this does and does not tell you. It tells you, honestly, that for one period neither side could impose itself: the demand that had been carrying an uptrend just met supply it could not overpower, or the selling that had been driving a decline just met a bid that would not break. That is real information, and on the right day it is early information. It does not tell you what happens next. A tie is a description of one finished period, not a forecast of the following one, and almost every mistake made with dojis begins by quietly confusing the two.

This is why four numbers are enough to argue about. The open and close together fix the verdict; the high and low together fix how hard that verdict was fought for; and a doji is simply the case where those two facts point in opposite directions, a wide, hard-fought range wrapped around no verdict at all. Every other candle resolves the tension between effort and result in favour of one side. The doji leaves it unresolved on purpose, which is why it feels like the chart pausing to ask a question, and why the honest response to a question is to wait for the answer rather than to guess it.

The precision question comes first, because it is where loose articles go wrong. The strictest definition of a doji wants the open and close identical to the paisa. On a liquid instrument that exact tie is uncommon, so the working definition used in practice is relative: the real body has to be a tiny fraction of the period's high-to-low range.

How tiny? Here is the honest answer: there is no universal cutoff, and any source that quotes one number as the definition is really describing its own scanner settings. Pattern-scanning software commonly treats a body under roughly 3 percent of the range as a doji; other published conventions stretch the allowance to 5 or even 10 percent; a stricter school insists on near-identity. Every one of those is a modelling choice, not a law of markets. What actually matters is what the ratio means: body divided by range measures how much of the session's fight survived to the close. A 40 paise body on a 24 rupee range is under 2 percent, an unambiguous doji. A body that is a third of the range is not a doji at all; it is a spinning top, a close cousin where indecision was real but one side still kept measurable ground.

The reason the working definition is a ratio, rather than a fixed rupee amount, is that it has to travel across price levels. A 50 paise body means something very different on a 20 rupee stock than on a 3,000 rupee one; dividing the body by the day's range strips the price level out and leaves a pure measure of how decisively the session failed to resolve. It is the same instinct that makes a percentage move more informative than an absolute one, and it is what lets a doji on a broad index and a doji on a small-cap be compared at all, despite the two trading hundreds of points apart.

The name itself carries the idea. The candlestick vocabulary that came west through Steve Nison's 1991 reference text is Japanese, and the term doji is usually glossed as "the same thing", or "the same matter", for a session whose open and close print at the same level. A competing folk story ties it to a colloquial word for a blunder, on the logic that a full session ending exactly where it began is a rare oddity. The gloss is debated; the mechanics in the table above are not.

All dojis share the hairline body. Where they differ is where that body sits inside the range, and that position is a precise record of which probe failed. Read the wicks as attempted journeys and the body as the address where the session finally settled, and the four standard variants stop being shapes to memorise. They become four different sentences about the same kind of standoff.

The standard doji is the plain case: modest wicks on both sides, the body near the middle, both sides probed and neither kept anything. The long-legged doji is the same balance made violent: very long wicks both ways, a wide two-sided battle in which both extremes were rejected. Volatility expanded while direction died, and that combination is why the long-legged version carries the most drama, and, as the sizing section will show, the widest stop. The dragonfly and the gravestone are the directional pair. A dragonfly closes at the top of a long lower wick: sellers drove price deep and buyers took back every rupee. A gravestone closes at the bottom of a long upper wick: buyers drove price high and sellers took back every rupee.

This is the core of the subject and the line that separates a chart reader from a shape spotter. A doji has no fixed meaning. Its meaning is its position in the structure that came before it. The identical candle, same body, same wicks, same volume, writes three different sentences in three different places, and the figure below draws exactly that: one doji, three locations, three readings.

After a long directional run, a doji is the first candle in which the trend failed to displace price. Every bar before it moved the market in the trend's favour; then a full period of the same pressure produced nothing. That is the earliest measurable sign of exhaustion: the side that owned the move spent a session's worth of effort and could not buy another rupee of progress. It is a warning, not a verdict. Trends routinely pause, print a doji and carry on. The doji only says the engine missed a beat; the next candle says whether it has actually stalled.

Inside a sideways range, the same candle is close to worthless. A range is already a standing tie between buyers and sellers, so a doji there is balance inside balance, one more coin flip in a sequence of coin flips. It confirms what the previous twenty candles already established and predicts nothing. Most of the dojis on any chart live here, which is the plain statistical reason that "trade every doji" fails as a rule: you would mostly be trading noise.

Ahead of a scheduled event, a results announcement, a policy decision, a budget, dojis appear for a duller reason still: participants step aside. Positions are squared, quotes thin out, and the candle goes flat not because a great battle balanced but because almost nobody showed up to fight. Reading exhaustion into that candle is reading conviction into an empty room. It is uncertainty, and uncertainty resolves with the event, not with the next candle.

Two refinements separate a careful reading from a lazy one. First, the longer and more one-sided the run into the doji, the more the candle is worth: a doji after three or four strong candles driving into a level says the dominant side just failed for the first time, while a doji after a weak, choppy drift is barely a whisper, because there was little conviction to exhaust in the first place. Second, dojis cluster. A single doji is one failed period; two or three in a row at the same level describe a market that has tried repeatedly to continue and repeatedly could not, a heavier statement than any one of them alone, though it still waits on the same confirming close before it means anything you can act on.

A doji records a tie, and nobody rational acts on a tie; they act on its resolution. Confirmation is the next candle closing beyond the doji's range: after an uptrend doji, a candle that closes below the doji's low; after a downtrend doji, one that closes above its high. That close is the market announcing that the argument the doji recorded has been settled, and in which direction. The figure below shows the two outcomes that matter, drawn from the same doji.

The mechanics are strict on two points, and the figure turns on both. First, it must be a close, not a touch. An intraperiod poke beyond the doji's range that closes back inside it is just another failed probe, the very rejection the doji itself already recorded, and it proves nothing. Second, the close must clear the whole range, not merely the body. The wicks are prices real participants paid during the fight, so a resolution that cannot even escape the fight's territory is not a resolution. The most convincing confirmations tend to be decisive, wide-bodied candles that close well beyond the range on rising participation; a bullish engulfing candle after a downtrend doji is a textbook example of the market answering the question with force.

Why not act on the doji alone and take the better price? Because the doji is symmetric. It says the trend failed to displace price for one period; it does not say the trend has ended, and stalls resolve in both directions about as often as not. Entering on the doji is paying for an answer the market has not given yet. The confirming close does cost you distance, since price has already moved by the time you act, and that cost is real. It is also the fee for evidence, and the framework below prices it into the position size rather than pretending it away.

Two dojis can be drawn identically and mean opposite things, and volume is what tells them apart. Remember what the candle claims: a full period of two-sided combat that ended even. That claim is only credible if there was combat. A doji printed on heavy volume is a genuine, expensive stalemate: a large crowd traded hard in both directions and neither side could move the close. A doji printed on thin volume is usually not a stalemate at all; it is an absence, a session where so few participants showed up that price simply drifted back to the open for lack of anyone to push it.

The distinction is practical, not decorative. The whole value of a doji at the end of a trend is the idea that the dominant side finally met resistance it could not overcome. Heavy volume is the evidence that it actually met something: real supply absorbed real demand, or the reverse. Light volume at the same spot more likely means the trend's participants simply took a day off, in which case there is nothing to exhaust and the trend often resumes the moment they return. This is why a doji into a holiday, a half-day session, or the dead hour before a major announcement deserves far less weight than its shape suggests. To go deeper on reading participation, see our full guide to volume in trading; here it is enough to treat volume as the credibility check on every doji you find.

The same effect scales across timeframes and instruments. A daily doji on a broad index such as the Nifty 50 compresses a whole 09:15 to 15:30 session of nationwide participation, so even a quiet one carries some weight. A one-minute doji on the same index compresses sixty seconds, much of which may be a single algorithm resting a quote. The shape is identical; the amount of real trading behind it is not, and it is the trading, not the shape, that you are actually trying to read.

If a confirmed doji is traded at all, the stop is not a matter of taste: it goes beyond the doji's far extreme. The entire reading rests on one claim, that a probe was made and fully rejected. If price later reclaims the whole of the doji's range, that claim is dead by definition, and the trade must not outlive the idea that justified it. Below the doji's low for a long, above its high for a short.

That rule has a consequence most articles never carry to its conclusion: the doji's range sets your position size. Work the rupee arithmetic on a stock near ₹500 after a multi-week decline into support (every figure here is illustrative). A daily long-legged doji prints: high ₹512, low ₹488, open ₹499.60, close ₹500.00, a 40 paise body on a 24 rupee range. The next day closes at ₹514, above the doji's high, which is confirmation. The long is taken near ₹514 with the stop just beyond ₹488, so the stop distance is roughly ₹26 a share. A trader risking ₹2,600 on the idea can size at 100 shares, no more. Now rerun it with a compact doji: range ₹496 to ₹504, confirmation close ₹506, stop just beyond ₹496, distance roughly ₹10. The same ₹2,600 risk budget now allows 260 shares. Identical setup logic, identical rupees at risk, wildly different size, and the only thing that changed was the width of the candle's fight.

Drama is width, width is stop distance, and stop distance is the inverse of position size. Size off the stop, never off the account.

This is the professional treatment of the long-legged doji in particular. Its drama is real, a huge two-way battle with no winner, but the trader who sizes off capital instead of off the stop ends up risking the most on the widest, most violent setups, which is exactly backwards. The checklist below is the whole discipline in four lines, and every line exists to keep a memorable shape from turning into an oversized position.

Because the shape is so recognisable, the doji attracts more confident misuse than almost any other candle. Four mistakes account for most of it, and each maps directly onto something covered above. Notice that none of them is a charting error; each is a discipline error, a decision taken a step too early or a step too loosely.

Every one of these has the same cure, which is also the single sentence this whole guide is built to earn: let position, confirmation and the stop, not the shape, decide what you do. A trader who internalises that treats the doji as a prompt to check three things rather than a button to press, and the difference between those two habits is most of the difference between reading a chart and merely recognising a picture on it.

Here is the part a page selling you certainty would leave out. Candlestick patterns have been tested by academics, carefully, and the results for patterns traded in isolation are poor. The landmark study is Marshall, Young and Rose (2006) in the Journal of Banking and Finance: candlestick signals, doji-based reversals included, tested on Dow Jones Industrial Average component stocks from 1992 to 2002 against a bootstrap null, that is, random price series built to share the same statistical properties. The strategies added no significant value. The same team then ran the harder test, Marshall, Young and Cahan (2008) in the Review of Quantitative Finance and Accounting: the 100 largest Tokyo Stock Exchange stocks from 1975 to 2004, candlestick charting examined in its own home market over three decades. No value in the full period, none in the sub-periods, none in bull markets or bear markets. Later work across other markets reads mixed at best, and the mixed results tend to involve added filters, holding-period choices and cost assumptions doing much of the lifting.

Read that finding precisely, because it is about a specific object: the isolated pattern, traded mechanically, every occurrence, no context, no confirmation, no sizing discipline. That object has little standalone edge, and any article promising that a doji "predicts reversals" is contradicted by the best evidence available. But notice what the finding is not. It is not a test of a trader who ignores mid-range dojis, demands a confirming close, stops out where the reading is falsified, and sizes off the stop. That framework exists because the naked pattern carries no authority. The doji's honest job is informational: it tells you where a genuine argument happened, which side's probe failed, and, through its extremes, the exact price at which your interpretation would be proven wrong. That last item, a built-in invalidation level, is worth more than any claimed predictive power, because it is the input that risk management actually runs on. That reading order, context before candle, confirmation before entry, size from the stop, is exactly what the method we teach is built around.

The practical discipline, then, fits in three sentences. When a doji prints, say the sentence it wrote: who probed, who was refused, and where in the structure this happened. If the position is meaningless, a mid-range candle or a pre-event blank, move on without regret. If the position is meaningful, do nothing until the market finishes the paragraph with a close beyond the range, and let the width of the candle, not the strength of your conviction, decide how large the trade is allowed to be. Do that consistently and the doji stops being a signal you obey and becomes a question you answer, which is the only honest way to use it.

• Marshall, Young and Rose (2006), Journal of Banking and Finance 30(8), 2303 to 2323. The benchmark academic test of candlestick trading strategies, doji reversals included, on Dow Jones Industrial Average component stocks, 1992 to 2002, using a bootstrap methodology; it found the strategies added no significant value. ideas.repec.org
• Marshall, Young and Cahan (2008), Review of Quantitative Finance and Accounting 31(2), 191 to 207. The follow-up test on the 100 largest Tokyo Stock Exchange stocks, 1975 to 2004, candlestick charting examined in its home market: no value in the full period, the sub-periods, or bull and bear regimes. link.springer.com
• Steve Nison, Japanese Candlestick Charting Techniques (1991). The reference text that carried candlestick vocabulary to Western markets; the source of the standard definitions of the doji family and of the usual "same matter" gloss on the term.
• Exchange session timings and scanner conventions. The NSE continuous equity session runs 09:15 to 15:30, the participation window a single daily candle compresses; pattern-scanner documentation is the source for the common 3 to 10 percent body-to-range conventions and for the fact that no universal cutoff exists.