FOREX & PROP FIRM INTELLIGENCEHow our reviews work

The rules, before the result

A view you can check.

We publish the direction first, preserve the evidence, then compare it with the next dated reference rate.

What the signal means

Each publication is a daily directional research view on one of EUR/USD, GBP/USD, USD/JPY, USD/CHF, AUD/USD or USD/CAD. Bullish means the first currency is expected to strengthen against the second; bearish means it is expected to weaken. Workers AI may issue a no-signal decision when it cannot justify a direction.

The model reviews the latest reference observation, changes over one, five and twenty observations, five- and twenty-observation averages, and recent Federal Reserve or ECB announcements available to our source collector. It writes a conditional thesis and a reason that thesis could fail. A separate model pass checks factual claims against the supplied evidence. These checks reduce unsupported claims; they do not validate predictive accuracy or guarantee an outcome.

The data and its limits

Price observations come from the European Central Bank’s daily reference rates, delivered through Frankfurter with the ECB provider explicitly selected, or the ECB’s own feed. We do not blend different providers. Cross-rates, percentage changes and simple averages are calculated by fx.reviews; they are not additional ECB forecasts.

The ECB usually publishes around 16:00 CET on working days, excluding TARGET closing days. These are information-only reference observations, not live executable quotes. We do not infer intraday highs, lows, order fills, stops or targets from them. The underlying data can be obtained free of charge from the linked sources.

When we publish

The Worker runs every day at 17:35 UTC (21:35 Dubai), with a second check at 19:35 UTC (23:35 Dubai). It first assesses outstanding signals, then generates at most one new view using that day’s complete rate basket. The later run retries delays and cannot replace a published view. Weekends and TARGET holidays produce no new duplicate signal.

If fresh data, sufficient history or the source review is unavailable, scheduled publication waits. We never backdate a prediction using an outcome that is already known. The archive begins with the first actual publication; it contains no invented historical wins.

An editor may publish an outlook before that day’s reference release using the previous ECB business day’s observation. Such a view displays the source date separately from its publication date. Its assessment remains the next ECB business day after publication, so the reference comparison spans more than one observation and can include movement that occurred before publication. It is not a measure of trading performance. The scheduled job cannot replace the early view.

The result labels

  • Confirmed: the assessment reference observation moved in the published direction by more than 0.01%.
  • Not confirmed: the move was more than 0.01% against the published direction.
  • Flat: the absolute move was 0.01% or less.
  • Awaiting data: the assessment reference observation has not yet been verified.
  • Unverified: the expected observation date has passed without a verified result. We keep checking that exact date.
  • No signal: the model published a neutral decision rather than a directional view.

The assessment date is fixed when the signal is published: the next ECB reference business day after publication, excluding weekends, New Year’s Day, Good Friday, Easter Monday, 1 May and 25–26 December. A Friday view normally checks against Monday, not Saturday. If an expected observation is missing, a later day’s rate cannot substitute for it.

The calculation is (assessment reference ÷ original reference − 1) × 100. We store the original published direction, reference, timestamp, sources and threshold. Outcome classification comes from code, not AI. AI adds a retrospective explanation after the comparison. If the recap fails review, the factual result and a calculated summary still appear.

Direction is not trading profit

Confirmed does not mean an executable trade succeeded. Reference observations may predate the publication time, and actual dealing prices, spreads, fees, slippage and financing can differ. We therefore show reference-rate changes and outcome counts, not a strategy win rate or claimed return. This is educational research, not personalised investment advice.

Broker links and paid placements

Broker directory links help readers continue their research. Affiliate links are labelled, and we may earn a commission if you use them. Any paid signals partner is separately labelled Sponsored. A placement does not mean a broker endorses the signal or offers identical prices or products in your jurisdiction.

Broker partners cannot choose the pair, direction, reference price or result. Signals placements require a separate agreement and are independent of the directory’s monthly listing. Discuss a signals partnership.

Method version: ecb-daily-direction-v1 · Introduced 8 September 2026. Report a source or calculation error.