Candlestick charts, explained properly

Start with what a market actually emits: a stream of trades, each with a price, a size and a timestamp, in no particular rhythm. Nothing in that stream is a candle. A candle is what you get when you agree on an interval — one minute, one hour, one day — and then summarise every trade that fell inside it with four numbers.

Open

The price of the first trade inside the interval.

High

The highest price traded at any point during it.

Low

The lowest price traded at any point during it.

Close

The price of the last trade before the interval ended.

Those four are measurements of one thing, not four prices in sequence. It is a common and costly misreading: people describe a candle as “it opened here, went up to here, came down to here and closed here”, which is a story about a path. The candle contains no path. High and low tell you the extremes were reached; they tell you nothing about the order in which they were reached, or how many times, or how long the price stayed there. One candle is compatible with an enormous number of different hours.

The rectangle is the body and it always runs between open and close. The thin line above and below is the wick (also called the shadow or the tail), and it runs out to the high and the low. Which end of the body is the open depends entirely on direction, which is why the colour matters: on a rising candle the open is the bottom of the body, and on a falling one it is the top.

Every chart has to decide where one interval stops and the next begins, and every serious data store makes the same choice: intervals are half-open. The 09:00 hourly bucket contains every trade from 09:00:00.000 up to but not including 10:00:00.000. A trade at exactly 10:00:00.000 belongs to the 10:00 bucket, not to both and not to neither.

That gives the bucket two candidate names — the time it opened and the time it closed — and the convention almost everywhere is to label it by its open. So the candle labelled 09:00 describes what happened between 09:00 and 10:00. It is worth being explicit about, because it is one of the few genuine incompatibilities between data sources: at least one major data API labels the same bucket by its closing instant, and reading it as an open-labelled series shifts every bar on the chart by exactly one interval.

Once you stop reading a candle as a path, the body and the wick become two separate pieces of information. The body answers where did this interval settle relative to where it started. The wick answers how far was price pushed and then rejected.

Wide body

Opened near one extreme and closed near the other. One direction dominated the whole interval.

Almost no body

Opened and closed at nearly the same price despite covering a range. Often called a doji. It says the interval was contested and ended where it started — not that nothing happened.

Long upper wick

Price traded well above the close and did not stay there. Buyers reached a level sellers were willing to defend.

Long lower wick

The mirror image: price was pushed down and recovered before the interval ended.

It is worth resisting the temptation to turn those descriptions into predictions. Every one of them is a statement about what already happened, and a shape that reliably preceded a move would not have stayed reliable for long once it was widely known. What the shapes are genuinely good for is reading the interval you are looking at without having to watch it happen — which, on a chart of two thousand bars, is the entire job.

Almost every chart colours candles by direction, but there are two definitions of “direction” in common use and they disagree on a meaningful fraction of bars:

Close vs open

The bar is green if it closed above its own open. This is what most crypto and FX charts do, and it is the convention on this page.

Close vs previous close

The bar is green if it closed above the previous bar’s close. Common on equity charts, where an overnight gap makes the previous close the more meaningful reference.

A bar that gaps down at the open and then rallies all interval is green under the first convention and red under the second. Neither is wrong; a chart that does not tell you which one it uses is a chart you cannot fully read. The same applies to hollow-versus-filled candles, which encode the second convention with the border colour and the body fill doing two different jobs at once.

Because a candle is an aggregate, changing the interval does not show you different data — it shows you the same data summarised more or less coarsely. And the fold is precisely four rules: open of the first, highest high, lowest low, close of the last. Volume, where a venue reports it, simply sums.

Look at what the fold destroys. The hourly view shows a push to 103.40, a failure, a flush to 98.40 and a recovery. The daily bar records the extremes and a close of 102.80 — a modest green bar with wicks on both sides. Both are accurate. The daily bar is not a summary that lost the truth; it is a different question being asked of the same trades.

The practical consequence is that “what interval should I use” is really “how much do I want the aggregation to hide”. A shorter interval shows more structure and more noise, and both of those grow together. There is no interval at which the noise disappears and the structure remains.

Eventually you will find a chart with a missing bar, and there are two entirely different reasons for it. Either nobody traded during that interval, or the data source omitted a bucket it had no trades for. Some venues emit a filler bar for a quiet interval — a zero-volume candle carrying the previous close as all four prices — and some simply skip the bucket. Both are defensible; they are just different, and a chart built on both without normalising them will show two different kinds of hole for the same underlying event.

That, and several other structural differences between data sources, is a subject of its own — it is why the same pair on two exchanges does not produce the same candles. The third guide in this series covers it.

When you open an unfamiliar chart, four questions settle almost everything about how to read it, and they are all answerable in a few seconds:

What interval?

Every shape on the chart is a property of the interval, not of the market.

What timezone?

Bucket boundaries move with it, and so does every candle’s shape near the boundary.

Which colour convention?

Close versus open, or close versus previous close.

Is the last bar closed?

If it is still forming, three of its four numbers can still change.