A figure is only useful once you can see where it came from.

Every price on this site traces back to the European Central Bank's euro reference rates. The ECB sets these rates each trading day at around 14:10 CET, and publishes them at around 16:00 CET on its own site.

The ECB describes these rates as produced for information purposes only, not for transacting. They are not dealing prices. A bank or trading firm quoting a client a minute after the fixing is already working from a different number, so any rate shown here is a record of where the market stood at that moment.

What we calculate from them

Pairs such as EUR/USD and USD/JPY are built from the ECB's euro crosses. Where a figure on this site is not the published rate itself, a percentage change, a typical day, or a range position among them, it is our own calculation from those rates, and it is labelled that way at the point it first appears.

Weekends and ECB holidays carry no fixing. A run of trading days on the site skips straight over the dates the ECB did not publish.

A move of the same size in pips does not mean the same thing on every pair. Twenty pips on USD/CHF on a quiet week is not the same as twenty pips on GBP/USD on the afternoon of a rate decision. The normal-day multiple puts a day's move against that pair's own recent behaviour, so a trader can read it against something other than a raw number.

How the typical day is worked out

For each pair, the daily change in pips across the previous 20 ECB fixings is taken, and the standard deviation of that set of 20 changes is calculated. That figure is the pair's typical day, the spread of daily moves that pair has shown under its own recent conditions.

What the multiple tells you

The normal-day multiple is the day's move divided by the typical day. A multiple of 1.0 sits inside what the pair has been doing lately. A multiple of 2.5 means the move was two and a half times the size of a typical day for that pair, over that 20-fixing window.

Reproducing it

1. Take the ECB reference rate for the pair across its most recent 20 fixings.
2. Work out the day-on-day change in pips for each of those 20 days.
3. Calculate the standard deviation of that set of 20 changes. That is the typical day.
4. Divide the move you are looking at, in pips, by the typical day to get its multiple.

This is the same calculation behind the typical day column on the pairs index and behind the Is this move unusual? tool, which returns a move's multiple once a pair and a move in pips are entered. The tool draws on a 20-fixing window fixed at the time it was built, so a move entered against a pair whose recent volatility has since shifted, or figures typed in against the wrong pair, will return a multiple that does not describe the market as it stands.

An analysis piece on this site runs in the same order every time: a finding, a plain version, the full technical account, sources, what would change the view, and corrections. The order does not change from one pair to the next or one event to the next, so a reader who wants the number can stop reading after the first line, and a reader who wants the mechanism knows exactly where to look.

1. Finding. One sentence stating what happened, with the figure and the date it covers.
2. Plain version. The mechanism behind the move, written without jargon, separate from the conclusion itself.
3. Full account. The technical version, with a chart, a timeline, and the figures that support the finding.
4. Sources. The publisher and date behind every figure used in the piece.
5. What would change the view. The data or event that would alter the reading, named in advance.
6. Corrections. A dated record of anything changed since publication, or a line confirming there is nothing to correct and when the piece was last checked.

Why the structure does not move

A reader checking a single pip figure against a source should not need to read a page of prose to find it. Keeping sources in the same position on every piece means a figure can be checked without searching for it, and a reader who disagrees with the mechanism can see what would need to be true for the finding to change.

No trade view

None of the six sections carries a trade idea or a signal. The structure reports what happened and what is scheduled against it, and stops there.

What counts as a correction

A wrong date, a mislabelled source, a figure that does not match the calculation shown next to it, or a number the European Central Bank itself revised after it was used here, all count as a correction. A new fixing replacing yesterday's in the ordinary run of the day does not; that is simply the next figure arriving on schedule.

How a correction is shown

A correction is added to the page it affects, not quietly swapped into the original text. The date it was made sits next to a short account of what changed and why, so a reader who checked the figure before the fix can see what was wrong and what replaced it.

The corrections record

Every analysis piece on Analysis carries a corrections section as part of its fixed structure. Where nothing has needed changing, the section says so, with the date the piece was last checked. Where something has, each entry is dated on its own and states what changed.