Public forecast records: preserve the original wording before reviewing the outcome

An outcome review needs the forecast as it originally appeared

A public forecast can look more accurate after its wording, threshold or deadline is edited. Before judging an outcome, preserve the exact version you intend to assess and the criterion stated with it. Put any later copy beside the original rather than replacing it. This worksheet makes that distinction visible for one record, without pretending that a local form can prove what was published in the past.

The report links a supplied original snapshot, its stated publication time, an event cutoff and a later assessment. It checks whether those entries are internally consistent. It does not decide whether the claim was true, whether a publisher changed a post or whether an external archive is authentic. Its usefulness is in exposing missing evidence before a performance claim is made.

Write the criterion independently of the outcome

A phrase such as likely to rise leaves many questions unanswered. Which value, which source, compared with what, at what time, and under what threshold rule? Preserve the criterion that was actually stated or agreed before the event. If it was missing, do not invent a precise rule after seeing the result and then treat the original prediction as if it contained it.

The fictional criterion uses the first published value of an invented series and a strict threshold above 100. Exactly 100 would not satisfy that rule. The tool does not parse that prose or independently compare a real published number with it; you enter a supplied assessment and evidence note. Its job is to preserve which wording the assessment claims to use, not to manufacture a score.

Separate publication, cutoff, observation and report time

The original publication time describes when you say the forecast existed. The cutoff describes the event boundary chosen for this record. The outcome observation time describes when the supplied assessment evidence was observed. The report as-of says when you are describing the set. These are four different fields, and a later observation cannot establish an earlier publication merely by pointing to the same URL.

For this narrow model, the stated original publication must be strictly before the cutoff and the outcome observation must be at or after it. A publication exactly on the boundary is not a pre-cutoff record. The tool rejects a supplied observation later than the report as-of. These rules fit this particular audit; they do not cover every forecast whose condition could be satisfied before its deadline.

The autumn repeated hour can reverse the apparent clock order

NPL explains that the UK removes the summer hour on the last Sunday in October, moving from BST back to UTC-based standard time. Around that transition, two different instants can share similar local clock labels. Comparing only 01:45 and 01:15 without their offsets can therefore reverse the chronology you intended to check.

The fictional date is 25 October 2026. Its stated publication at 01:45+01:00 is 00:45 UTC. The cutoff at 01:15Z is 01:15 in London after the change. Although 01:45 looks later than 01:15 on a clock face, the publication instant is 30 minutes earlier. The worksheet keeps both original strings and London offsets, and it rejects unqualified local times rather than guessing which occurrence you meant.

Sources: National Physical Laboratory — When do the clocks change?.

The later copy is different even when the original assessment is usable

The default original says above 100; the later copy says above 90. Those strings differ, so the report clearly flags the later copy. The supplied assessment still refers to the original snapshot and original criterion, and its fictional note says 101. Under the entered chronology and references, the record is structurally usable, while both versions remain preserved.

Now change the assessment basis to later_copy. The original forecast can no longer be assessed under this report’s rule, so the usable supplied assessment becomes unknown and the reason is shown. The tool does not silently substitute the easier threshold. Conversely, a spelling-only edit can also produce an exact-text difference; the report does not claim every changed character changes the forecast’s meaning.

An entered URL and timestamp are references, not authentication

The original snapshot needs a saved reference and a source URL. The outcome assessment needs its own reference, URL, observation time and note. Missing any of these leaves the corresponding evidence incomplete. A valid HTTPS address is not a confirmation that the page existed at the claimed time, and this tool never opens the address.

For a serious external review, the provenance and timestamping method must be examined outside this worksheet. Preserve the original source file or archive reference and explain how its date was established. A later screenshot, a current page or a self-entered date may be useful clues, but the form cannot turn them into a verified pre-event publication.

Recording a plan is different from creating a registration

The OSF dictionary distinguishes ordinary projects from registrations and preregistrations. It describes registering a research plan before reviewing data, with timestamped updates and visible changes. That is an example of a documented evidence process with its own rules. This worksheet does not submit anything to OSF or reproduce that service’s controls.

The downloaded CSV remains editable. No cryptographic signature, trusted timestamp, public post, independent receipt or immutable audit trail is created here. Saving the file is a way to keep your current evidence together, not proof of historical priority. If another system supplies a registration or archive receipt, preserve its exact reference without claiming that this page independently validated it.

Sources: Center for Open Science — OSF Dictionary.

Passing the cutoff does not turn an unknown into a failed forecast

The report can show that its as-of time is after the cutoff while the outcome remains unknown. That may mean the observation is missing, the criterion is unclear, or the evidence has not been reviewed. The tool does not label an absent outcome as not_met, and it does not create a win rate from a selected list of completed examples.

An outcome timestamp before this model’s cutoff is flagged as unsupported for the required post-boundary assessment. A missing original publication time likewise leaves pre-event status unknown. These are evidence gaps, not accusations of dishonesty. Preserve the unresolved fields and update a new report when the necessary information becomes available.

Keep the original snapshot attached to every conclusion

Read the record state, original-time ordering, assessment basis and later-text comparison together. A true statement about one field cannot repair another missing piece. In particular, a well-ordered timestamp does not authenticate its source, and an unchanged copy does not prove the criterion was clear or the outcome correctly assessed.

The export includes the original text, later copy, original criterion, references, all entered times and every unresolved issue. Editing invalidates the previous output, and manual copying remains available when clipboard access is unavailable. With JavaScript off, this method and the fictional timing table remain readable. The optional link to the site’s separate offer is not required for the record review.

Worked examples

{
  "forecast_id": "DEMO-FORECAST-001",
  "original_text": "DEMO: the first published value for invented event E will be above 100.",
  "later_text": "DEMO: the first published value for invented event E will be above 90.",
  "original_criterion": "DEMO original criterion: first published value above 100 means met; 100 or below means not met. Keep later revisions separate.",
  "published_at": "2026-10-25T01:45:00+01:00",
  "original_reference": "DEMO saved original snapshot reference",
  "original_url": "https://example.org/demo-forecast-original",
  "cutoff_at": "2026-10-25T01:15:00Z",
  "as_of": "2026-10-25T01:30:00Z",
  "outcome_at": "2026-10-25T01:20:00Z",
  "outcome_reference": "DEMO saved outcome reference",
  "outcome_url": "https://example.org/demo-outcome",
  "outcome_state": "met",
  "assessment_basis": "original_snapshot",
  "outcome_note": "DEMO record says 101; reader marks the unchanged original criterion as met. The worksheet has not verified this value."
}

Questions about the method

Primary sources and scope

1. National Physical Laboratory — When do the clocks change?Checked 2026-10-20. UK BST/GMT transition and last-Sunday-in-October rule support the repeated-hour example. No publication timestamp is verified.
2. Center for Open Science — OSF DictionaryChecked 2026-10-20. Distinguishes plans, registrations and preregistrations, including timestamped updates. No OSF account, registration or immutable evidence is created by this tool.

Examples explain a method. They do not predict prices or establish where a trading service is available. Check the named source before relying on future schedules.