No. 03

Method card

How a forecast becomes a record

The rules that decide when an ARCANE Call counts, and who is allowed to say it came true.

Result
0
Outcome
Method
Updated
25 Sep 2026
Fig. 1 · Drawn from the published numbers
HELD-OUT ENROLMENT · 25 NOV – 25 DEC 2026REVIEW NO EARLIERTHAN 25 JAN 2027NOV 2026DECJAN 2027FEBAT LEAST 30 INDEPENDENTLY RESOLVED CASES BEFORE ANY RESULTEACH SQUARE IS ONE CASE · EMPTY UNTIL THE WINDOW RUNSENROLMENT · 25 NOV – 25 DECREVIEW ≥ 25 JANNOV 2026DECJAN 2027FEB30 RESOLVED CASES BEFORE ANY RESULTEMPTY UNTIL THE WINDOW RUNS
Fig. 1The first prospective window for rates and credit Calls, as planned. Hatched is held-out enrolment, 25 November to 25 December 2026; the rule is the earliest review, 25 January 2027. Each empty square is one of the 30 independently resolved cases required before any result. Planned, not run.

Summary

0resolutions a model can approve on its own

A Call is scored only if it was recorded before its outcome. A model may propose a resolution, but only an editor can approve one, and only against an exact passage quoted from the source registered with the Call. Records are appended, never edited or deleted. Ambiguity stays open; it never becomes a NO.

What a Call is

Every ARCANE article carries a read: a thesis, the observable that would show it, a horizon, what would confirm it and what would break it. A Call is that read in a form that can be checked later against an exact source and deadline.

How a Call resolves

The source registered with the Call is retrieved again and hashed. A model may read it and propose an outcome, but it must quote the exact passage it relies on. A passing quote only moves the Call to editor review.

The editor checks the passage against the original criterion and deadline, including whether it reports an observation or only another forecast. Before anything is written, the source is retrieved once more and must match the reviewed copy byte for byte. The decision is appended once. A changed page or criterion needs a fresh review.

What we will not do

We do not create a historical forecast, choose a missing deadline, or change an outcome once recorded. The absence of confirming text is not a NO. Every attempt, refusal and unresolved case is kept, so the record cannot be improved by quietly dropping the misses.

What comes next

A first prospective window for rates and credit Calls is planned, with held-out enrollment from 25 November to 25 December 2026 and review no earlier than 25 January 2027, and only with at least 30 independently resolved cases. These are planned windows, not evidence that anything has run. Public timestamping of each run is under evaluation and not yet switched on.

Limitations

  • Source passages and receipts are private ledger material; the method is public, the retained copies are not redistributed.
  • PDF and other binary sources stay unresolved until a checked text path exists.
  • An editor is still a single point of judgement. The quoted passage is kept so any reader can second-guess it.

What would change this

Any one of these would change the conclusion

  1. Any recorded Call found to have been created or altered after its outcome was known.

Source boundary

Published
the resolution rules and the planned evaluation windows.
Withheld
storage, retention policies, internal schemas and source copies.

Changelog

  1. Resolution custody rules adopted.
  2. First public edition on Labs.
  3. A figure drawn from the published numbers added. No wording or number changed.

Cite

@techreport{arcane2026howa,
  title       = {How a forecast becomes a record},
  author      = {{ARCANE Research}},
  institution = {ARCANE Intel},
  type        = {Method card},
  year        = {2026},
  url         = {https://arcaneintel.net/research/how-a-forecast-becomes-a-record}
}