Foundations · Lesson 3 of 8 · 16 min read

Reading candlesticks

A candlestick is a lossy compression of a trade tape, and knowing exactly what it lost is most of what it takes to read one. This lesson folds one real minute into one bar, then takes the bar apart again.

Assumes you have read: How an order book works
After this you can: Say exactly what a candle recorded, and what it does not claim about what happens next.

The short version

A candle is four numbers taken from every trade inside one time bucket: the first price, the highest, the lowest and the last. Nothing else survives.
Two minutes that went in opposite directions can produce one identical candle. The bar records the endpoints and the range, never the route.
Changing the interval does not change the market, only the size of the bucket. The same trades, grouped differently, draw a different picture — and both pictures are correct.
The rightmost candle on a live chart has not happened yet. Its close is just the latest print, and its high and low can still extend until the interval ends.

Four numbers out of a thousand trades

The first two lessons dealt in things that exist: a book of resting orders, and the instructions that consume it. Every time one of those instructions crosses the spread, a trade prints to the tape — a price, a size and a timestamp. A busy minute on BTC/USDT produces hundreds of them.

A chart cannot draw hundreds of prints per minute across a day and stay legible, so it throws almost all of them away. Take every trade inside one interval and keep four numbers: the open (the first price), the high (the largest), the low (the smallest) and the close (the last). Draw those four as a shape and you have a candle.

A candle is not a summary of the trades. It is the trades, with everything except four of their numbers deleted.

That framing is worth holding onto, because almost every mistake people make with candles is a mistake about what was deleted. The four numbers are exact — they are real prices that really traded, not averages or estimates. Everything between them is gone.

The fold, in fifteen lines

Here is one minute of BTC/USDT, reduced to eight prints so it fits on a page. A real minute has far more; the reduction changes the density, not the rule.

Eight prints and the bar they fold into. The dashed guides are the reduction happening — first print to the body's lower edge, last print to the upper edge, and the two extremes straight across to the wick tips. Everything between the guides is discarded.

Swipe horizontally to compare all columns.

Eight trade prints inside the minute beginning 12:04:00 UTC, and which of them the candle keeps
Time · UTC	Price	Size · BTC	Kept as
12:04:03.112	64,298.4	0.084	first print — this is the open
12:04:11.980	64,305.0	0.211	—
12:04:19.402	64,318.7	0.045	—
12:04:27.155	64,326.2	0.630	highest price touched — the high
12:04:38.740	64,309.5	0.118	—
12:04:44.010	64,292.8	0.372	lowest price touched — the low
12:04:52.663	64,301.9	0.096	—
12:04:58.501	64,311.3	0.240	last print — this is the close
The candle	O 64,298.4 · H 64,326.2 · L 64,292.8 · C 64,311.3 — volume 1.796 BTC

Worked example — fold the eight prints by hand

Run the rule down the table one row at a time. The state starts empty; each print either moves one of the four numbers or moves nothing but the close.

1 · 64,298.4O H L C = 64,298.4A new 60-second bucket. All four numbers start at the first price.
2 · 64,305.0H, C → 64,305.0
3 · 64,318.7H, C → 64,318.7
4 · 64,326.2H, C → 64,326.2
5 · 64,309.5C → 64,309.5Below the high and above the low, so only the close moves. Most prints in a real minute do exactly this and change nothing else.
6 · 64,292.8L, C → 64,292.8
7 · 64,301.9C → 64,301.9
8 · 64,311.3C → 64,311.3

The finished bar64,298.4 → 64,311.3Open to close, having touched 64,326.2 and 64,292.8 on the way. The close is not a summary — it is simply whichever price happened to be last before the clock rolled over.

Notice how little work the rule does. Only four of the eight prints moved an extreme; the other three after the first just dragged the close along behind them. In a real minute of several hundred prints the proportion is far more lopsided than that — a candle is cheap to compute and lossy by construction, and those are the same fact.

How this product actually builds that bar

The worker that holds our venue sockets folds the 1m candle from the trade tape itself, one print at a time. The whole rule is: work out which 60-second bucket the trade belongs to; if it is a new bucket, start a bar whose open, high, low and close are all this price; otherwise raise the high if this price is higher, lower the low if it is lower, set the close to this price, and add the size to the volume.

That is not an approximation of a candle. A candle is the open, extremes and close of the trades inside its window, so folding the same prints the venue would fold produces the same bar. It exists because @kline_1m is one of the streams the futures venue acknowledges, reports as active, and then never sends — so the bar you watch forming on a perpetual is one we rebuilt from the tape rather than one we were given.

Body, wicks and colour

The four numbers become a shape with three parts, and each part has a literal reading.

Body — open to close. Filled green when the close is above the open.
Reversed body — the candle on the right closed below its open. Colour encodes only that comparison.
Wicks — the thin line reaching past the body to the highest and lowest prices touched inside the interval.
Measured — range 33.40 high to low, body 12.90, upper wick 14.90, lower wick 5.60. The drawing is to scale, so you can check that against it.

The worked minute drawn to scale: high 64,326.2 at the top, low 64,292.8 at the bottom, and the open and close landing exactly where they fall between them. Beside it, a candle whose body is reversed because it closed below its open.

The body spans open to close. It is where the interval settled its business. Our minute opened at 64,298.4 and closed at 64,311.3, so the body is 12.90 wide.
The wicks (or shadows) reach from the body to the highest and lowest prices touched at any point inside the interval. The upper wick here is 14.90 tall, the lower 5.60 — price went further up than it did down, and did not stay.
Colour encodes one comparison and nothing else: close above open is drawn up-tinted, close below open down-tinted. A green candle inside a week-long decline is still green. It is not a verdict.

The proportions matter more than the colour. A long body with short wicks says the interval moved in one direction and mostly stayed there. A long upper wick says price ventured much higher and was sold back before the close — buyers pushed, sellers answered, and the answer held. A long lower wick is the mirror. A candle whose open and close are almost equal — a doji — records an interval that went places and ended where it started.

One thing the shape never tells you is when inside the interval each of those things happened. That is the subject of the next section, and it is the single most useful fact about candlesticks.

What the candle threw away

Four numbers preserve the endpoints and the full vertical range of an interval. They throw away the route between them entirely — and the route is where most of what people believe they are reading actually lives.

Two minutes that went in opposite directions and produce one identical bar. A sold off first and rallied into the close; B rallied first and was sold back down. Both open at 64,298.4, touch 64,326.2 and 64,292.8, and close at 64,311.3.

These two minutes have opposite stories and the same candle. Any sentence of the form “this candle shows sellers stepping in near the top” is a sentence about the route, and the route is precisely what the bar does not contain. It might be true. The candle is not the evidence.

There is one honest way to recover some of it: drop to a smaller interval. Every candle is made of smaller candles, and the five 1m bars inside a 5m bar carry more of the path than the 5m bar does. That is not a trick — it is the same trades, bucketed finer. It is also why professional charts are read at two intervals at once rather than one.

The interval is a zoom, not a truth

Our chart offers eight intervals: 1m, 5m, 15m, 30m, 1h, 4h, 1d and 1w. Switching between them does not fetch a different market. It fetches the same trades in a different bucket size.

Five one-minute candles and the five-minute candle they fold into. Open comes from the first minute, close from the last, and the extremes from wherever in the run they happened — here the low was set in the middle.

Swipe horizontally to compare all columns.

Five one-minute candles and the five-minute candle they aggregate into
Minute · UTC	Open	High	Low	Close
12:00 up	64,280.0	64,296.4	64,278.9	64,295.0
12:01 up	64,295.0	64,318.9	64,293.4	64,316.2
12:02 down	64,316.2	64,320.4	64,276.1	64,304.7
12:03 down	64,304.7	64,309.8	64,288.5	64,298.4
12:04 up	64,298.4	64,326.2	64,292.8	64,311.3
12:00 – 12:04 up	64,280.0	64,326.2	64,276.1	64,311.3

Read the last row against the five above it. The 5m open is the 12:00 open, unchanged. The 5m close is the 12:04 close, unchanged. The high came from 12:04 and the low from 12:02 — the middle of the run — because extremes belong to whichever bucket contained them. Nothing is averaged and nothing is smoothed.

Two of those five minutes closed below their opens. The bar they add up to is green. That is not a contradiction and it is not noise: a bucket has one open and one close, and what happened to the four boundaries inside it is exactly the information the aggregation deletes.

So “which interval is right?” is not a question with an answer. The interval is the question you are asking. A 1m chart asks what is happening now; a 1d chart asks what this month has been. They disagree constantly, and both are telling the truth.

The rightmost candle is not finished

Every candle on the chart except one is a closed record of a completed interval. The one on the right edge is still being written. Its open is fixed, its close is merely the latest print, and its high and low can still extend in either direction until the bucket ends.

This catches everyone once. A 1h candle that looks like a decisive green bar at minute fifty can be a red one with a long upper wick at minute sixty, having recorded nothing new except a few more trades. Nothing changed about the past; the bucket simply had not closed.

Candle timestamps on this chart are the venue’s UTC bucket-open times, and the crosshair readout labels them UTC for that reason. A chart that silently renders venue time in your local zone puts every bar on the wrong clock, and the error is invisible until you try to line a candle up against something that happened at a known moment.

Volume: the other half of the bar

Underneath the candles our chart draws volume: the total size that traded inside each interval, tinted to match its candle’s direction. In the worked minute it is 1.796 BTC — the eight sizes in the table, added up.

Volume is the one thing on the chart that is not a price, and it answers a different question: not how far the market moved, but how much had to change hands to move it. A large move on heavy volume involved a lot of participants. The same move on thin volume was cheap to cause — which usually means it is cheap to undo.

This is where the first lesson pays off. Volume is the tape, aggregated. Depth is the book, aggregated. A price that ran a long way on very little volume ran through a thin book, and the ladder would have shown you that before it happened.

Patterns, honestly

There is a large folklore of named candle formations — hammers, engulfing pairs, morning stars — each with a prophecy attached. It is worth being clear about what a pattern can and cannot be, because this is where reading candles turns into believing things.

A pattern is a shape made of the four numbers. Since the four numbers are a real record, a shape is a real description: a hammer genuinely records an interval that fell hard and closed near its high. That description is worth having. The prophecy attached to it — that the next interval rises — is a claim about the future that the shape does not contain, made on a sample the folklore never publishes.

The honest position is the one this whole lesson has been building toward. A candle is evidence about what happened. Two candles are evidence about what happened twice. Neither becomes a forecast by being given a name.

Do this now, it takes three minutes

Open the BTC/USDT chart and leave it on 1m. Watch the rightmost candle for one minute — you will see its close move with every print and its wicks extend, then the bar finish and a new one start on the next boundary. That is foldTrade running.

Now find a candle with a long upper wick and switch the interval one step finer. The wick becomes several candles, and you can see which part of the interval it happened in — the information the coarser bar deleted, recovered exactly the way this lesson said it could be.

Then look left at the book and the trade tape beside the chart. The tape is the prints; the chart is those prints in buckets; the book is what is left waiting. Three renderings of one machine, and candles stop being decoration the moment you see that.

Five ways to misread a candle

Mistaken belief: “The long upper wick means sellers stepped in at the top.”
What actually happens: It means price traded that high and did not close there. Whether that was one large seller, a slow drift back, or a spike and an instant reversal is exactly the information the candle deleted. Drop an interval and look, or say “I do not know” — those are the two honest options.
Mistaken belief: “This 1h candle is green, so the hour was bullish.”
What actually happens: It means the last print of the hour was above the first. Two of the five minutes in the aggregation table above closed red inside a green 5m bar, and the same arithmetic scales: a green hour can contain forty red minutes. Colour is a comparison of two endpoints, not a description of an interval.
Mistaken belief: “The candle is still forming but the shape is clear.”
What actually happens: The shape of an unfinished candle is a shape of an unfinished interval, and it can reverse entirely before the bucket closes. If a decision depends on what a bar looks like, it depends on the bar being closed — otherwise you are reading a sentence someone is still typing.
Mistaken belief: “The 1m and the 1h disagree, so one of them is wrong.”
What actually happens: Both are correct renderings of the same trades at different bucket sizes. They are answering different questions, and the disagreement is the answer to a third question: how much of the recent move is inside the noise of the larger frame.
Mistaken belief: “Volume confirms the move.”
What actually happens: Volume measures how much traded, and every trade has a buyer and a seller, so it never has a direction of its own. High volume says a lot of people participated in whatever the price did; it does not say they agreed with it. The tape’s aggressor colouring, from lesson one, is the closest thing to a direction — and it is a much narrower claim than “confirmation”.

Check yourself

Five questions against the worked minute — open 64,298.4, high 64,326.2, low 64,292.8, close 64,311.3. Answer before you open each one.

Did the price touch 64,320.0 during this minute?

Yes. It is between the low and the high, and the wick reaches 64,326.2, so every price in that range traded at some point. What you cannot say is when, how often, or in which direction it was crossed.

Was the price above 64,320.0 at the halfway point of the minute?

Unknowable from the candle. The bar records four numbers and three of them are extremes with no timestamp. This is the same question the two-paths figure answers with “both, depending which minute you were in”.

The 5m candle containing this minute is green. Does that make this minute green too?

No, and the reverse is equally untrue. This minute happens to be green as well, but two of the five minutes in the aggregation table are red inside the same green 5m bar. Each bucket compares its own first and last print and nothing else.

You are watching a 1h chart and the rightmost candle has not moved in ten minutes. Is the feed broken?

Probably not. Our live bar is applied on the 1m interval only, because painting a 1m bar onto a 4h chart would draw a candle that does not exist. On 1h the last bar advances when the history reloads. To check the feed is alive, look at the trade tape or the book — or the status page, which says which mode each feed is in.

Two candles have identical bodies, but one has twice the volume. What does that tell you?

That the same net price change required twice as much size to happen. It says nothing about direction — every unit of volume had a buyer and a seller. It is a statement about how much participation the move needed, which is a real thing to know and a much smaller thing than “confirmation”.