What is a doji candle?
A doji is a candle whose open and close finish at or very near the same price, leaving a body that is a thin line rather than a block, with wicks that may extend in either or both directions.
The textbook interpretation is indecision: buyers and sellers fought through the period and finished where they started. That reading is reasonable as a description of what happened inside the bar and considerably weaker as a claim about what happens next.
Several named variants exist. The long-legged doji has extended wicks both ways; the dragonfly has a long lower wick and almost none above; the gravestone is its mirror. The distinctions are real geometry, and whether they behave differently is a measurable question rather than a settled one.
There is also a definitional wrinkle worth naming, because it determines your entire sample. 'Open and close are equal' is almost never exactly true, so every doji rule needs a tolerance — body smaller than some fraction of the total range, or smaller than some fraction of recent average range. Change that tolerance and you change how many dojis exist by an order of magnitude.
Do doji candles signal a reversal?
Taken alone, a doji is one of the weakest reversal signals available, because it occurs far too often to be selective and because the same shape appears in trends, ranges and dead periods alike.
The arithmetic is the argument. On many instruments and timeframes, candles meeting a reasonable doji tolerance make up a meaningful share of all bars. A signal that fires on a large fraction of bars cannot separate much of anything; whatever happens next after a doji is close to whatever happens next in general.
What changes the picture is location. A doji at a level that already mattered — a swing high, a flip zone, the edge of a range — is a different proposition from a doji in the middle of nowhere, because the level is doing the selection and the candle is only supplying timing.
That is the practical reframing worth taking away. Candles are timing tools, levels are location tools, and a candle without a location is a signal with no selectivity. Testing a doji rule in isolation usually produces a result indistinguishable from the base rate, and this is not a failure of the test.
How do you test a doji strategy properly?
Test it by comparing what happened after doji bars against what happened after all bars in the same period, rather than by looking only at outcomes following dojis.
This comparison against a base rate is the whole discipline for any candle pattern. If price rose over the next ten bars 52% of the time following a doji, that means nothing until you know it rose 52% of the time following any bar. The interesting quantity is the difference, and the difference is usually far smaller than the headline.
Fix the tolerance before you start, and count the variants if you try more than one. Body under 5% of range, under 10%, under 0.1 ATR — each is a separate trial, and the most attractive of six thresholds is a search result rather than a discovery.
Split by context rather than pooling. Doji at a level versus doji in open space; doji in a trend versus in a range; doji at the session open versus mid-session. Pooling averages a real effect in one context with its absence in others, and the average tends toward nothing — which is exactly what most published candle research reports.
Where do doji candles actually help?
Dojis are most useful as a confirmation step inside a setup that already has a reason, supplying the moment of entry rather than the reason for it.
A concrete example: price approaches a flip zone that your testing says is worth trading, and you need a trigger to enter rather than a limit order sitting in the zone. A doji forming at the zone edge is a defensible trigger because it marks a pause in the approach, and the candle's own extreme gives you a natural invalidation point.
Used that way, the doji is not being asked to carry any weight of its own. It is being asked to answer 'now?' after something else answered 'here?', and that division of labour is the difference between a testable structure and a chart-reading habit.
The corresponding warning is about timeframe mismatch. A doji on the timeframe you execute on can be a completely ordinary bar on the timeframe your level came from, so requiring the candle on the wrong timeframe adds delay without adding evidence.