Why a Signal Disappears Before the Bar Closes

The bar on the right has not finished happening

The right-most bar on any chart is still being built. Its open is fixed, because the period has started. Its high, its low and its close all change with every quote that arrives until the period ends.

So whatever shape it has at the moment you look at it is a snapshot of an unfinished period, and the finished period can be a different shape entirely: a long body that ends up short, a wick that is taken back, a close that lands on the other side of the open. Nothing has malfunctioned when that happens. The reading was always of something incomplete.

Everything drawn on top of it inherits the same status

Because those numbers are provisional, so is anything computed from them. A study whose input includes the forming bar moves as the bar moves. A named shape whose final leg is the forming bar can complete and then un-complete. A level that appears to have been touched may have been touched only by a wick that no longer exists once the period closes.

This is the ordinary explanation for the complaint that a signal vanished. Nothing was removed and no data was revised. The period simply finished differently from the way it was going, and every layer built on the period followed it.

Closed bars or live: a setting, not a mystery

Most tools can read either the last closed bar or the one in progress, and which they read changes what the same tool reports on the same chart at the same instant. It is normally a documented option rather than hidden behaviour, and it is the first thing to check when two people running an identical study disagree about whether something is on screen.

Alerts deserve checking separately from the study that feeds them. A tool can compute on closed bars and still fire the moment price touches a level intrabar, or compute continuously and only notify at a close. The two settings are independent, and knowing one does not tell you the other.

Two very different things get called repainting

The first is what has been described so far: a tool that reads the forming bar and changes with it. That is honest, and it resolves the moment the period closes. There is nothing to fix and nothing being concealed.

The second is a tool whose output over the past changes as new bars arrive, because its definition requires bars that come after the point it marks. That is not dishonest either, but it is a different claim, and it has a consequence worth being plain about: the historical chart is showing marks that could not have been made at the time they appear to refer to. Which of the two you have is answered by the tool’s own definition, and the two are not interchangeable.

A pivot only exists in hindsight

The clearest case is the turning point. A high becomes a swing high because the bars on either side of it failed to exceed it, and how many bars must fail is a parameter somebody set. At the moment the high is made, nothing distinguishes it from a bar that will be exceeded on the next one.

A chart of the past is therefore covered in clearly marked pivots that nobody could have marked at the time. Anything that reads them off history — a line drawn between two of them, a shape defined by them, a rule that counts them — is consuming information the live chart did not contain. That is not an argument against looking at history. It is an argument for knowing which parts of a historical chart were available live and which were not.

Confirmation trades one cost for another

Waiting for a period to close replaces a provisional reading with a settled one. That is not free: the settled reading arrives later, and later means at whatever price exists then rather than the one that produced the provisional signal. Reading the forming bar is not free either, since some of what it shows will not survive the close.

There is no version of this without a cost, and the size of the cost is set by the time frame — one bar is a few minutes on one chart and a month on another. What matters when reading anyone’s description of a method, including your own, is whether it says which of the two it took. A description that does not say has left out the part where the difficulty lives.

What a chart image does not show

Almost every chart published to illustrate a point is an image of closed bars, captured after the periods resolved. The same chart while it was live contained a bar in motion, and often a shape that had not yet become the shape in the picture.

This is why an annotated screenshot can be entirely accurate and still not demonstrate what it appears to demonstrate. It establishes that the thing was visible afterwards, not that it was visible at the time. The question that separates the two is simply when the image was captured relative to the bars it marks — a fair question to ask of any chart offered as evidence, including one you produced yourself.