Why the Same Pair Looks Different on Two Platforms

There is no official chart of a currency pair

Spot forex has no central exchange and no consolidated record of what was traded. Prices are quoted by banks and other liquidity providers to whoever they deal with, and no venue exists whose job is to publish the resulting transactions in one place. Every chart you have ever looked at was therefore built by somebody — a broker, a platform vendor, a data provider — out of the price sources it happens to have, using rules it chose. Two of them will not agree exactly, and neither is a corrupted copy of an original, because there is no original. Almost every disagreement described below follows from that single structural fact.

Which side of the quote the bars are drawn from

A pair has two prices at any moment, the bid and the ask, separated by the spread. A bar has room for one of them, so the provider decides which. Building from the bid is the common retail convention; building from the ask, or from the midpoint of the two, are ordinary alternatives, and some platforms expose it as a setting.

The consequence is that the whole chart shifts vertically by roughly the width of the spread depending on the answer. A level drawn on one basis does not sit at the same number on another, a high or a low can print on one and not on the other, and the difference grows exactly when the spread does — in thin hours and around scheduled releases. Which basis a chart uses is stated in the platform’s own documentation, and it is worth knowing before comparing your screen with anybody else’s.

Whose prices went into the bar

Even two providers both drawing bid bars are aggregating different inputs. A broker streaming its own book is showing the prices it could deal at, drawn from the liquidity providers it has relationships with. A third-party charting service is showing a composite of whatever sources it subscribes to. The two overlap without matching.

They diverge most at the extremes, which is the awkward part. An outlying quote that one aggregate contains and the other does not becomes the high or the low of a bar in one and not in the other. So the disagreement is largest at precisely the values that levels, patterns and protective orders are read from, and smallest in the middle of the range where it would not have mattered.

Where the day begins

A daily bar is defined by a clock, and the clock belongs to the platform. The server’s time zone decides the moment one day’s bar closes and the next opens, so two platforms whose servers sit in different zones produce genuinely different daily opens, highs, lows and closes for the same market. Everything computed on daily bars inherits that difference, including any study whose window is counted in days.

Daylight-saving transitions move the boundary during the year. The same applies further up: where the week begins decides which bar carries the reopening, and how a provider treats a holiday session decides whether a bar exists there at all. None of this is an error on anyone’s part. It is a convention, and conventions differ.

What the scale does to a line you drew

The vertical axis is a rendering choice with consequences. On a linear scale, equal height means an equal amount of movement; on a logarithmic scale, equal height means an equal percentage change, so a move is drawn according to the level it started from. The prices plotted are identical either way.

What changes is anything drawn by hand across them. A trendline joining the same two points is a different line on each scale, and a break of it can be present on one and absent on the other. The effect is negligible over a narrow range and grows with the span covered, which is why it surfaces on long histories and stays invisible on short ones. Automatic scaling adds a milder version of the same problem, since the apparent steepness of everything depends on how much of the chart is on screen.

How far back the history goes, and where it came from

Stored history is a property of the platform rather than of the market. How many bars a chart holds decides how far a long-window study can be plotted at all, which is why some indicators only begin part-way along a freshly opened chart. It also decides what a backtest can even be run over.

The question worth asking, and answerable only from a provider’s own documentation, is whether the older part of a chart was built by the same process that draws the live part. A history assembled one way and streamed another is one chart in appearance and two in construction. The practical version is narrower: if a study you rely on is computed over a long window, most of its input is the platform’s stored bars rather than anything you watched arrive.

What travels between platforms, and what does not

A level, a named shape, a screenshot, a backtest result: none of these are properties of a currency pair. Each is a property of a chart, and a chart is the output of four decisions somebody made — which provider, which side of the quote, which session boundary, which scale. Two people can look at the same pair, disagree about whether a level held, and both be reading their screens correctly.

That is not a reason to distrust charts. It is a reason to state those four settings whenever a reading is meant to be checked by anyone else, and to treat a reading that omits them as unreproducible rather than merely contested. All four are documented by the platform, which is the only authority available — because in this market no other one exists.