The methodology behind every call. Nothing hidden. Everything auditable.
EdgeLetter is a continuous monitoring system that detects geopolitical developments, scores them by confidence, publishes signals with timestamps, and grades its own outcomes against independent evidence. Every step of this process is documented here. If the archive says something was verified, this page explains exactly what that means.
Why publish this at all? Because a system that does not explain how it grades itself cannot be trusted. Every claim on this page corresponds directly to how the pipeline actually operates. If something on the archive does not match this methodology, that is a bug — not a feature.
01 / Signal Detection
How EdgeLetter identifies developments before the market recognises them.
The pipeline runs continuously, ingesting sources across geopolitical, military, trade, and macroeconomic domains. When a development is detected, it enters a scoring process before any decision is made about publication.
1
Source ingestion
The pipeline monitors live sources across geopolitical risk, military activity, trade and sanctions, energy and commodities, monetary policy, and corporate intelligence. Sources are ingested continuously, not on a scheduled crawl.
2
Extraction and structuring
Each development is extracted into a structured format: headline, what happened, why it matters, what changed, the mechanism, and invalidation conditions. This structure is preserved in the published signal and in the Protection Card if the signal is later verified.
3
Confidence scoring
Every candidate development is scored 0–100 for confidence. The score reflects source strength, corroboration, specificity of the claim, and how directly it relates to a monitored thesis. Only developments scoring 55 or above are published. Everything below 55 is held on the watchlist or rejected.
PUBLISH THRESHOLD: 55/100 Below 55 → Watchlist (hold) or Rejected 55 and above → Published signal
4
Publication with timestamp
When a signal is published, its timestamp is recorded permanently. This timestamp is the reference point for all lead-time measurements. It cannot be changed after publication. The timestamp is what makes the archive auditable.
5
Subscriber delivery
Published signals are delivered via the Morning Brief, the subscriber dashboard, and direct alerts. Delivery is tiered: free accounts receive headlines, Starter and above receive full signal bodies, Professional and above receive WhatsApp delivery and invalidation conditions.
02 / Scoring System
What the confidence score means — and what happens at each threshold.
The confidence score is not a prediction of whether something will happen. It is a measure of how strongly the available evidence supports the signal at the moment of publication. A score of 72 means the evidence is strong. A score of 41 means it is weakening. A score below 40 means the thesis may be breaking.
Score
Status
What it means
72 – 100
PUBLISHED · HOLDING
Strong evidence. Signal published or thesis actively holding. No action required unless invalidation conditions are triggered.
55 – 71
PUBLISHED · STABLE
Evidence supports publication. Signal crosses the threshold. Watchlist items in this range are approaching publication.
40 – 54
WATCHLIST · WEAKENING
Below the publish threshold. Being monitored. If evidence strengthens it may publish; if it continues weakening it will be rejected.
0 – 39
WATCHLIST · BREAKING
Thesis may be collapsing. Subscribers receive an alert when a previously published signal's assumption drops below 40.
The watchlist is the pre-publication layer — assumptions the system is monitoring but has not yet determined are strong enough to publish. Subscribers see the watchlist on their dashboard as conviction scores in real time. This is the earliest possible view of a developing situation.
03 / Recognition Doctrine — RECOG_v3
Exactly what "verified" means. No exceptions.
Verification is the most important word in the EdgeLetter archive. When a Protection Card says VERIFIED, it means one specific thing, defined by the RECOG_v3 doctrine. Not "widely reported." Not "broadly believed." The exact conditions below, and nothing else.
RECOG_v3 Recognition Formula
VERIFIED= second independent corroborating source ANDEvent AND ( ThemeORStrongLocation ) AND ( ActorORStrongLocation )
What each term means:
Event — A specific, dateable occurrence must be confirmed. "Tensions are rising" is not an Event. "Iran reimposed vessel restrictions on 18 April" is an Event.
Theme — A domain classification: Energy & Commodities, Geopolitical Risk, Trade & Sanctions, Defence & Security, etc. Named strategic chokepoints satisfy Theme when paired with an Event.
StrongLocation — A named geographic chokepoint with established strategic significance. These satisfy either Theme or Actor requirements when an Event is confirmed.
Actor — A named state, institution, or non-state actor whose actions are confirmed by the second source.
Second independent source — The corroborating source must be editorially independent of the first source. Two articles from the same publisher do not satisfy this requirement.
Named chokepoints that qualify as StrongLocation:
Strait of HormuzSuez CanalRed SeaBab-el-MandebStrait of MalaccaTaiwan StraitSouth China SeaBlack SeaBosphorus
04 / Lead Time Measurement
How the gap between EdgeLetter and the market is measured.
Lead time is the interval between when EdgeLetter published a signal and when the second independent mainstream publisher confirmed the same development. It is computed automatically from timestamps. It is not estimated, adjusted, or curated after the fact.
EdgeLetter publishesT=0
Second source confirmsT + lead
Lead time = T+lead − T=0 · Computed from timestamps · Not adjusted retroactively
1
EdgeLetter publication timestamp (T=0)
The exact UTC timestamp when the signal was published to the platform. This is the start of the measurement. It is stored at publication and cannot be modified.
2
Recognition timestamp (T+lead)
The publication timestamp of the second independent mainstream source to confirm the same event. For the RECOG_v3 doctrine to be satisfied, this source must confirm the Event AND satisfy the Theme/Actor/StrongLocation requirements.
3
Lead time calculation
Lead time = Recognition timestamp − EdgeLetter publication timestamp. Expressed in hours to one decimal place. If lead time is under one hour, expressed in minutes. Negative lead time (EdgeLetter later than the market) is published as-is — not hidden.
4
Evidence links
Every Protection Card includes the URLs of the evidence sources used to determine the recognition timestamp. Anyone can independently verify the lead time calculation by checking the publication dates of those sources against the EdgeLetter timestamp.
05 / Outcome Grading
How every signal is graded — including the ones that were wrong.
After a signal's review window closes, the system grades the outcome against what actually happened. This grading is automated and not subject to human editorial judgment. The grades are assigned by the system and published permanently. They cannot be changed.
CORRECT
Thesis held
The development EdgeLetter signalled occurred and was independently confirmed. The lead time represents a genuine window of decision time that subscribers had before the market recognised the event.
Protection delivered: subscriber was ahead of the market.
INVALIDATED
Thesis broken by a subsequent event
A subsequent development overtook the original signal — often a ceasefire, policy reversal, or geopolitical shift that changed the assumptions the signal was built on. The lead time is still valid: subscribers had advance warning of the assumption break.
Protection delivered: subscriber knew the thesis was breaking before the market did.
PARTIALLY CORRECT
Thesis partially confirmed
Some but not all of the signalled development occurred or was confirmed. The direction was correct; the magnitude or specific mechanism differed from what was signalled.
INCORRECT
Thesis did not hold
The signalled development did not occur or was contradicted by subsequent evidence. These cards are published with the same visibility as Correct cards. A system that hides its incorrect calls cannot be trusted.
AMBIGUOUS
Insufficient evidence to grade
The review window closed without sufficient independent evidence to determine whether the thesis held or not. The signal is preserved in the archive with this grade rather than forced into a binary outcome.
NOT APPLICABLE
Grading not relevant
Applied to monitoring signals, structural developments, or ongoing situations where a binary correct/incorrect grade would misrepresent the nature of the intelligence.
Intelligence Utility Rate. EdgeLetter publishes two accuracy metrics. Forecast Accuracy counts only Correct and Partially Correct outcomes. Intelligence Utility Rate also credits Invalidated outcomes — because detecting an assumption break early is a form of protection, regardless of whether the original thesis held. A subscriber who knew 6.3 days early that a ceasefire was coming was protected, even though the original escalation signal was overtaken by events.
06 / The Archive
What goes in, what can never be changed, and why.
The Protection Archive is the permanent record of every signal EdgeLetter has published. The rules governing what can and cannot happen to a Protection Card are fixed, not policies — they are how the system is built.
🔒
Protection Cards are immutable after freezing
Once a Protection Card is created from a published signal, its headline, timestamp, lead time, evidence links, and outcome grade cannot be edited. The system stores them as insert-only records. There is no update path.
👁
All cards are publicly visible
Every Protection Card — including Incorrect and Ambiguous outcomes — is publicly accessible at edgeletter.pro/proof-archive. Subscribers do not need to be logged in to audit the complete record. Anyone can check.
⏳
Timestamps are recorded at publication, not retroactively
The EdgeLetter publication timestamp is stored at the moment a signal crosses the publish threshold. It is not assigned after verification, after the event occurs, or after any subsequent review. The timestamp is the timestamp.
📊
Evidence links are stored with each card
The sources used to determine the recognition timestamp are stored on every verified Protection Card. If a link goes dead, the timestamp and the domain are still on record. The verification can still be independently assessed.
❌
Incorrect outcomes are never removed
The Protection Archive contains EdgeLetter's worst calls alongside its best. This is not a marketing decision — it is a design requirement. A selective archive proves nothing. A complete archive, including failures, is the only kind that can be trusted.
The methodology is only as credible as the archive that proves it.
Every claim on this page corresponds to a real mechanism in the pipeline. The best way to verify it is to open the archive, pick any Protection Card, and check the timestamp against the evidence sources.