Editorial policy

This page is the set of rules a number has to survive before it appears on this site, and what happens to it when it turns out to be wrong. It is written down because a rule nobody can read is a rule nobody can hold us to.

Last reviewed 11 October 2026

What has to be true before a figure is published

It came off the rig, or it says where else it came from three source classes, printed at the number: our measurement, a third party, the broker's own terms
Its window is printed with it a figure without the window it was measured in is not published
Its series passed the feed audit a series that looks like a simulator rather than a market disqualifies the broker from ranking
Nothing is ranked across classes a broker we measured and a broker quoting its own price list are not in one column
No source class, no number the build refuses to publish a figure whose class nobody declared
Missing is missing an axis we did not measure prints no data and the reason, never a dash and never an estimate

What we do not publish at all

No star ratings, no scores out of ten, no “best broker” verdicts and no top-ten lists. Those are opinions with a number stuck on them, and the moment one appears on a site funded by a broker it is worth nothing. We also do not reprint a broker's claimed withdrawal time, support speed or execution quality as a finding — if we have not measured it, the axis says so.

We do not publish a figure from an account type we did not open, we do not average across account types at one broker, and we do not fill a gap in a window by interpolation. A month with a hole in it is reported as a month with a hole in it.

Corrections

A wrong number is corrected openly. The faulty rows are flagged and excluded from every published figure rather than deleted — a tick cannot be collected again, and a record of what the rig got wrong is worth keeping because the next fault usually rhymes with the last one. The corrected figure then carries the same window and method as before, and what changed is recorded.

How a correction runs

ReportedBy a reader, a broker, or our own watchdog. One address: desk@costcheckfxen.com.

CheckedAgainst the raw rows, not against the published page. If the raw rows are gone, the figure comes down rather than being defended.

QuarantinedBad rows are flagged and excluded everywhere, including in past windows.

RepublishedWith the corrected value, the same stated method, and a note of what changed.

Not doneQuietly editing a number and hoping nobody kept the old page.

Right of reply

A broker that believes a figure about it is wrong can write to the desk and will get the raw query and the rows behind that figure. If the figure is wrong we correct it. If we still think it is right after looking, the broker's reply is published next to the figure it concerns, unedited except for length, and the reader can decide. What a broker cannot do is have a correct measurement removed, and no amount of commercial relationship changes that — including the one with the broker that pays for this site.

We also apply the rule to ourselves. Our own sponsor's feed has already failed our audit once and was excluded from ranking as a result; that is in the methodology, with what it cost us. A desk that only ever finds faults in other people's data is not auditing anything.

Nothing on this site is advice, and you can lose money trading. Figures describe a stated past window on accounts we opened ourselves.