High confidence · profile
Inputs are current, the broader read is well supported, and the user can treat the setup as a stronger research candidate.
Triggered when
Drivers agree · data is fresh · regime is stable.
Methodology
Neural Edge outputs a decomposed market read, research priority, directional bias, confidence, driver context, and the conditions that could weaken the thesis. Built for analyst workflows, not standalone execution. Every read exposes its drivers, freshness, uncertainty, and opposing evidence.
What we measure
The same five families run on FX, metals, and equity indices, grouped into interpretable context, not collapsed into an opaque number.
The public surface is explanation-first. It exposes the families doing the work, so the reader can judge the case rather than trust the score.
Real yields, policy path, and central-bank trajectory drive every asset class, FX rate differentials, gold’s real-rate sensitivity, and the equity discount-rate channel. Neural Edge exposes whether the ranking is being led by durable macro context or something more fragile.
Rates · Policy path · Real yields · Central banks.
Cyclical residuals against consensus shape FX growth differentials, equity earnings revisions, and the metals demand backdrop. The same lens, surfaced differently per asset class.
Growth · Inflation · Earnings · Demand.
CFTC, dealer flow, ETF flows, breadth. Crowdedness is a risk signal across all three asset classes, it qualifies the macro read rather than replacing it.
Positioning · Flows · Breadth · Crowding.
Realised against implied vols across tenors. Risk-appetite shifts move FX, metals, and equity in correlated ways, a good macro read can still be a poor workflow fit if conditions are unstable.
Volatility · Regime · Risk appetite.
Dollar strength, terms-of-trade, liquidity stress, scheduled catalysts. Surfaced explicitly per asset class, never hidden inside the score. Pulse adds what changed, what it affects, and how urgent.
USD strength · Liquidity · Calendar · Headlines.
Five families. The same lens, surfaced differently per asset class.
Trust state
Confidence is trust in the current read, not directional certainty. A high score with low confidence should be read very differently from a high score with high confidence, and both stay visible.
High confidence · profile
Inputs are current, the broader read is well supported, and the user can treat the setup as a stronger research candidate.
Triggered when
Drivers agree · data is fresh · regime is stable.
Moderate confidence · profile
The read is useful, but one or more parts of the picture require more caution, cross-checking, or waiting.
Triggered when
Mixed evidence · some inputs aging.
Low confidence · profile
The setup is tentative, freshness-limited, or otherwise incomplete. It remains visible so uncertainty is explicit rather than hidden.
Triggered when
Conflict · stale inputs · event risk pending.
A neutral read is a real read. Confidence rises when drivers agree, data is fresh, and the regime is stable. It falls when evidence conflicts, inputs are stale, or major event risk can invalidate the read.
News overlay · Pulse
Pulse maps admitted headlines to the instruments they affect with a bounded, time-decaying overlay. It contextualises the Sovereign read; it does not re-rank the board on its own.
Every admitted headline is mapped to a bounded impact so a single story cannot dominate the ranking. Direction and magnitude stay explicit.
Impact decays on a published schedule so old headlines lose weight naturally rather than persisting as stale bias.
Headlines route to affected instruments through explicit mappings, per FX pair, per metal, per index. Pulse publishes instrument-level relevance with provenance.
Sovereign owns ranking. Pulse adds news relevance and urgency around it. Responsibility stays clear.
Pulse adds context, it never owns the score. Sovereign keeps ranking authority; Pulse adds news relevance and urgency around it. Engine boundaries stay visible so responsibility stays clear.
Sentiment indices conflate noise with signal and are heavily backward-looking. We score the underlying drivers, not the chatter about them.
Social chatter is not a primary proof layer in the product. The workflow is built around ranked context and explicit market drivers instead.
Retail flow is not promoted as a public decision layer. If a signal cannot be explained clearly inside the workflow, it does not become the headline promise.
The model has no view on chart shapes. Patterns that survive falsification become positioning or vol-regime inputs, not standalone scores.
Generic crude or commodity baskets are poor proxies for any individual instrument. We use the relevant complex per instrument (see Factor 05).
What we won't measure shapes what we will. Discipline about exclusions is what keeps the public surface honest, every absence is a decision, not a gap.
Public disclosure
This page publishes interpretation rules for the public product story. Live parameters, performance claims, and roadmap promises stay where they belong, inside the product and the doctrine.
Public surfaces use a signed magnitude with a separate bias field for direction.
Direction should not be hidden inside the sign of the score.
Confidence is shown as High, Moderate, or Low and expresses trust in the read, not directional certainty.
Confidence helps interpretation; it should not become marketing theatre.
Freshness and uncertainty belong in the workflow. The product makes stale or incomplete context visible rather than hiding it.
Users need to know when the read is current and when caution is warranted.
Pulse adds relevance and urgency context around headlines. It does not take over score ownership from Sovereign.
Engine boundaries matter in both the product and the public story.
The live product covers FX, metals, and equity indices. Stocks and crypto are deferred until the same factor discipline applies.
Scope discipline matters more than breadth theatre.
This page explains interpretation rules for the public product story rather than publishing a live parameter dump.
The public site stays current, bounded, and easy to read without pretending to be an internal research notebook.
Interpretation rules, not a live parameter dump. The public methodology stays readable, bounded, and current.