What a timeframe changes

How one candle is built out of smaller ones 

A timeframe is the length of time one candle on the chart covers. Nothing else about the chart changes when it is switched. The instrument is the same, the price history underneath is the same, and the axes mean exactly what they meant before. The only thing that changes is how much of that history is compressed into each mark, and compression is not a neutral operation.

The compression follows one rule, and it is arithmetic rather than judgement. Every candle of the lower timeframe that falls inside the higher timeframe's window is collected. The open of the aggregated candle is the open of the first of them. The close is the close of the last. The high is the highest high any of them reached, and the low is the lowest low any of them reached. Nothing is weighted, averaged, smoothed or interpolated, and no fifth value is consulted. That is the entire operation, and it is why two providers holding identical price data draw identical candles when their windows begin at the same instant, and different candles when the windows do not line up.

The hourly candle opens and closes at the same price. Read on its own it reports an hour in which price finished where it started, having travelled a little above and rather further below. The four candles underneath report a rise, a further rise, a fall through both of them, and a partial recovery. Neither account is wrong and neither is more accurate. The hourly candle answered the question it was asked, which was where the hour began, where it ended, and how far price reached on either side. The order in which it reached those places was never part of the question.

What each step up deletes 

Three distinct things are discarded every time the timeframe rises, and they are worth naming separately because each is lost for a different reason and each matters to a different kind of description.

• Sequence. A candle records that its high and its low both occurred somewhere inside its window. It does not record which came first. Two windows containing opposite internal stories can produce candles that are identical in all four values.
• Frequency. A candle records the extreme price reached, not how many times price arrived there. A level visited once and a level visited repeatedly leave the same single mark on the aggregated candle.
• Duration. A candle records where price went, not how long it stayed. Minutes spent at one price and seconds spent at another are drawn with the same ink, because the vertical axis carries price and the horizontal axis carries the window, not the time spent within it.

A candle close is a boundary, not an event 

Nothing happens in a market when a candle closes. No participant acts on it, no obligation falls due, and the instrument is unaware that a chart exists. The boundary is a property of the charting software's clock, and it follows that a candle close is defined by three things that have nothing to do with price: the timeframe selected, the time zone the platform's server keeps, and whether that zone observes daylight saving.

For a market that trades continuously across sessions this has a visible consequence. A daily candle drawn on a server that rolls the day at one hour is a different object from a daily candle drawn on a server that rolls it at another. The two charts describe the same market and disagree about every daily open, every daily close, and occasionally about the high, when a spike falls on one side of the boundary rather than the other. Practitioners disagree about which daily boundary is the correct one, and no authority settles it, because a market without a closing auction has no closing print for a convention to defer to.

Why a level appears on one timeframe and not another 

Resolution is the smallest movement a chart can render as a separate feature. A move that fills most of a five minute candle occupies a sliver of a weekly one, and on the weekly chart it is not a small feature. It is not a feature at all, because it has been absorbed into a mark that is drawn as one continuous object.

Two separate mechanisms make a price area appear on one chart and vanish on another. The first is the arithmetic already described. A turning point that is the extreme of an hourly candle need not be the extreme of the day containing it, so the point anchoring a description on one chart simply does not exist as a point on the other. The second is rendering. A chart has a finite number of pixels across its price axis, and two prices closer together than one pixel cannot be drawn as two separate things, however many times price turned between them.

The consequence runs through every lesson that follows. A description of price behaviour is always a description at a stated resolution, never a property of the price history alone. Two people looking at the same instrument, one on a fifteen minute chart and one on a daily chart, can describe genuinely different structures with neither of them misreading anything. Later lessons in this module name those structures and the levels they rest on. What this lesson establishes is that the name attaches to a pairing of a price history with a timeframe, and that quoting one without the other leaves the description incomplete.

The number of observations falls as the timeframe rises 

One year of history contains a fixed quantity of time, so dividing it into longer windows necessarily yields fewer candles. The arithmetic is unremarkable and its consequences are not.

Two consequences follow. The first is accumulation. Any description that requires a number of examples takes proportionally longer to accumulate on a high timeframe, so a formation occurring a handful of times a year on a weekly chart and one occurring many times a week on a five minute chart are not comparable as bodies of evidence, whatever they look like. The second is overlap. The candles on a lower timeframe are not additional history. They are the same history rendered finer, so counting the same formation on two timeframes counts the same underlying events twice, and treating the two counts as independent observations double counts them.

Where practitioners disagree 

The most common claim made about timeframes is that higher ones are less noisy. Half of that claim is not in dispute: aggregation removes variation by construction, so a higher timeframe chart has fewer marks and smoother contours as a matter of arithmetic. The contested half is the word noise, which is a judgement rather than a measurement. One tradition treats movement inside a candle as random variation that obscures structure, and regards aggregation as filtering. Another argues that noise is simply the name given to information the chosen resolution cannot represent, and that the same operation is therefore the removal of evidence rather than the removal of error. Both descriptions are consistent with the identical chart, which is why the argument has not resolved and will not resolve from the chart alone.

A second disagreement concerns whether time is the right basis for a bar at all. Time based candles give every window equal width, so an hour of dense activity and an hour of near silence are drawn as objects of the same size. Alternative aggregations close a bar after a fixed number of ticks, a fixed traded volume, or a fixed price range, and their advocates argue that the clock is arbitrary while activity is not. Their critics answer that bars built on activity are not comparable between sessions or between instruments, that they differ between providers because the underlying tick counts differ, and that no published reference price, settlement or scheduled economic release aligns to them. Time based candles are the convention on most platforms, and a convention is a convention rather than a finding.

A third disagreement concerns the unfinished candle. One convention treats only completed candles as observations, on the grounds that a forming candle's values are provisional and can reverse before the window ends. Another reads the forming candle, on the grounds that waiting for a boundary defined by a server clock is deferring to an event that happens in software rather than in the market. Neither position can be settled by the chart, for the reason set out above: the boundary is not a property of the instrument. Which timeframe a given reader works at is not something this page can assess, and it does not put forward one.