The usual definition of a confirmation candle fits in one line: the candle after a pattern that closes in the direction the pattern implies. That line is correct and almost useless, because it says what the candle looks like and nothing about what it is for. Most pages give you the line and then move on to indicators. This post stays on the candle, because it is doing a specific job, and once you can see the job you can decide for yourself whether to wait for it.
What is a pattern before it is confirmed?
A candlestick pattern is a description of what buyers and sellers did in one session, or two, or three. A hammer says: sellers pushed price well below the open, and buyers took all of it back before the close. That is a real event and it is worth noticing. But notice what it is not. It is one session’s behaviour, read after the session has finished, by you.
Everything you know about that hammer comes from the data that produced it. The read (“selling was tested here and failed, so demand is at this level”) is a hypothesis built from a sample of one, and a hypothesis cannot be tested on the data it was built from. If you have ever fitted a strategy to the past and watched it fall apart the next month, you already know this problem at a larger scale.
The next candle is the first out-of-sample test. Between the hammer’s close and the next close, every other participant who was watching the same chart saw the same hammer and had the chance to act on it, or not. The next candle records what they did. It is the only information about the pattern that did not exist when you first formed the read.
That is the whole mechanism. Confirmation is not a rule that makes a pattern true. It is the earliest moment the market gets a chance to disagree with you.
Why does it have to be a close and not a wick?
Because a wick is a price that was reached and not held, and the question confirmation asks is precisely whether the read held.
Take the hammer again. Suppose the next session pushes above the hammer’s high in the morning and finishes back inside its range. On a daily chart that prints as an upper wick through the level and a close below it. Someone will call that confirmed. It is the opposite. Buyers tried the level, sellers were waiting there, and the session ended with the sellers still in charge of it. A wick through a level is evidence that the level was contested, not that it was won.
A close is different in kind. It is the price the session ended on, after everyone who wanted to trade that day had traded. On a liquid daily chart the close is the closest thing the market has to a settled opinion. The same logic is why the body of a candle matters more than its wicks in the first place: as StockCharts puts it, candlesticks only reflect the relationship between the open and the close, not the sequence of events in between. The close is the part of that relationship that survived.
So the confirmation candle is defined by its close, and the close has to be past a specific price: the level the pattern itself proposed. For most single-candle reversals that is the pattern’s own high or low.
What confirms the hammer, the bullish engulfing and the morning star?
The three most-called bullish reversals, stated concretely. Each one is drawn to its real proportions and has a full page in the candlestick pattern library.
The hammer
- Confirmed by: the next candle closing above the hammer’s high.
- Invalidated by: a close below the hammer’s low.
The hammer is the pattern that most needs the test, because the shape is common and the evidence behind it is thin. In Bulkowski’s counts, the hammer acts as a reversal 60% of the time, which he describes as not far from random. That is not an argument against the hammer. It is an argument that an unconfirmed hammer is a coin flip with a slight lean, and the next close is where the lean either shows up or does not. For how the wider published tests of candlestick patterns compare, and why their results disagree, see does technical analysis work.
The bullish engulfing
- Confirmed by: a close above the engulfing candle’s high.
- Invalidated by: a close below the engulfing candle’s low.
The bullish engulfing is a stronger single event than the hammer, because one session undid the previous one outright. Bulkowski counts it as a bullish reversal 63% of the time, and warns in the same breath that the performance after the breakout can be dreadful. He catalogues the engulfing plus a third candle that closes higher as a separate pattern, the three outside up, which reverses 75% of the time in his data.
Read that gap carefully before you get excited about it. The three outside up is by construction the subset of engulfings where the next candle went up, so part of the difference is selection, not magic. What the figures do show is the direction of the mechanism: the third candle carries information the first two did not.
The morning star
- Confirmed by: the third candle closing past the midpoint of the first, which is most of the pattern’s own definition.
- Invalidated by: a close below the low of the middle candle.
The morning star largely carries its own confirmation, and this is the reason three-candle patterns are taken more seriously than single candles. The middle candle is the stall; the third candle is the test of the stall, already inside the pattern. Bulkowski has it reversing 78% of the time and treats a close above the top of the pattern as the breakout. If you want an extra test beyond the pattern, that is the one to use, but you are now waiting for a fourth close and paying for it. A break that closes past the level and then falls back inside it is the subject of false breakouts.
What does waiting actually cost?
This is the part the one-line definition hides, so here it is in illustrative numbers. None of these are real prices.
Say a hammer forms after a decline with a low of 96.00, a close of 99.00 and a high of 100.00. The invalidation for the read is a close below 96.00 either way.
- Acting at the hammer’s close means an entry around 99.00, with 3.00 between the entry and the invalidation.
- Waiting for confirmation means the next candle has to close above 100.00. Suppose it closes at 102.50. The entry is now 3.50 worse, and the distance to the same invalidation has more than doubled to 6.50.
That 3.50 is the price of the test. It is real, it is paid on every confirmed read, and no amount of framing makes it go away. Whatever move follows, a piece of it has already happened by the time you have your evidence.
Now the other bill. Acting at the hammer’s close means acting on every hammer, including the ones where the next session closes back below the low. On Bulkowski’s figure that is roughly four hammers in ten, and you find out which ones after you are in. Waiting removes many of those. It also removes some real reads: the hammer that confirms a day late, the one that gaps so far above the high that the confirmed entry is far worse than 102.50, the one that grinds sideways for three sessions and then goes.
Confirmation is a filter. A filter removes false reads and real ones together, and a worse entry is the fee. Not waiting keeps the better entry and the full error rate. Neither choice is the correct one; they are two different bills, and which one is cheaper depends on the instrument, the timeframe and how many false reads you can carry before you stop trusting the chart. This is what each choice buys, not a recommendation of either.
Where does the invalidation come from?
From the same moment. This is the part almost nobody writes down, and it is more important than the confirmation itself.
The instant you name the confirmation level, you have also named the level that proves the read wrong. For the hammer it is “confirmed above 100.00, wrong below 96.00” (still illustrative). Both numbers come from the same candle, and both exist before the next candle prints. A read that has a confirmation level and no invalidation level is not a read; it is a hope with a trigger attached.
The reason to write both down in advance is that the next candle can only test a claim made before it arrived. If you decide what “wrong” looks like after the candle closes, you will move the goalposts to wherever price is, every time, and the out-of-sample test never actually happens.
Some traders, once a read confirms, move the invalidation up to the confirmation candle’s low, on the grounds that the market has now shown a higher price it would not give back. That is a legitimate read and also a tighter, easier-to-hit level. Say which one you are using before the next candle, not after.
What can confirmation not do?
Two things, and both matter more than the definition.
It cannot put a pattern in the right place. A hammer only means something after a decline, because “selling was tested and failed” is only news where selling had been working. A hammer in the middle of a range, followed by a green close above its high, is a range doing what ranges do: price going up because it was recently down. The confirmation candle tests the read. If the read had no premise, there was nothing to test, and a green candle after a green candle is not evidence of anything. Check the context first; the bullish patterns roundup lists what each one needs to have happened before it.
It cannot turn one print into a vote. On a liquid daily chart, a close is the result of a full session’s auction, with a lot of participants and a lot of money settling on a price. On a 1-minute chart, a close is whoever happened to be there for sixty seconds. On a thin instrument, a daily close can be a single order. The confirmation candle draws its meaning from how many people had the chance to disagree with the read and did not. Where that number is small, the candle has the right shape and none of the meaning. It is just the next candle.
The read, in one line
A candlestick pattern is a hypothesis about buyers and sellers, formed from the session that produced it. The confirmation candle is the first session that could not have been used to form the hypothesis, which is why its close, and only its close, counts as evidence. Waiting for it costs part of the move; not waiting costs the false reads it would have removed. Either way, the read is not finished until it says both “confirmed above X” and “wrong below Y”, with real numbers, before the next candle prints.
If you want that stated on the chart in front of you, that is how Chartix reads a screenshot: the pattern, whether its context requirement is actually met, the confirmation level and the invalidation level, as prices. It will not tell you whether to wait. That part is yours.
Sources
Every figure above was checked against these before publishing.
- Bulkowski, Hammer
- Bulkowski, Bullish Engulfing
- Bulkowski, Three Outside Up
- Bulkowski, Morning Star
- StockCharts ChartSchool, Introduction to Candlesticks