Loading market data…
Markets

Methodology

A step-by-step, honest walkthrough of how a headline becomes a published MarketMood article with a Market Impact label.

1

Ingestion — Google News RSS

Our news-automation pipeline polls public Google News RSS feeds on a fixed cadence (roughly every 60 seconds) for headlines matching our tracked instruments (Gold, Silver, Bitcoin, Nasdaq, S&P 500, Crude Oil), major forex pairs, and Trump/White House economic-policy news. Junk aggregator/chart-page titles are filtered out.

2

AI/LLM classification

Each candidate headline is classified by an LLM (via OpenRouter) into Bullish or Bearish for the relevant instrument — Neutral-classified items are excluded from the public feed by policy. The model also generates a short, hedged-language reasoning summary (e.g. “potentially bullish,” “could pressure,” “may support”) grounded in the source headline — it is instructed not to make absolute or certain claims.

3

Importance heuristic

A separate, rules-based heuristic looks at signal strength (e.g. how directly the story ties to a tracked instrument, recency, and language cues) to flag a subset of stories as higher-importance/“Breaking.” This is an automated heuristic, not a human editorial desk ranking every story.

4

Publish, with source attribution

Approved articles publish with the original source name and (when available) a direct link back to the original reporting. Articles auto-expire from "current" surfaces (72h for general coverage, 24h for Breaking News) but remain permanently reachable at their own URL for transparency and backlink integrity.

5

Human review & correction

A human admin can review the live post list at any time, pin/lower/delete any article, or manually add a post the automated pipeline missed. This is the human-in-the-loop safety net on top of the automated pipeline — not a claim of full-time editorial staff pre-screening every single story before it goes live.

The per-instrument Market Mood / fundamental bias engine

The Live Signals bias shown on the homepage and category pages is generated per instrument from our own most-recent published coverage for that instrument, plus recent price momentum as a secondary signal, recomputed roughly every 60 seconds. Like the article-level labels, this is an estimate synthesized with AI assistance — not a certified trading signal.

The Market Mood Index (overall 0-100 score)

The Market Mood Index shown at the top of the homepage is a single sitewide number derived directly and transparently from the SAME six per-instrument biases described above (Gold, Silver, Bitcoin, Nasdaq, S&P 500, Crude Oil) — it is not a separate model with its own hidden inputs.

1

Score each instrument

Each instrument's current bias is converted to a number: Bullish = 100, Neutral = 50, Bearish = 0. This is the simplest honest numeric encoding of the three real states already computed by the bias engine — nothing more elaborate is hidden underneath it.

2

Average for the overall score

The overall Market Mood Index is a plain average of all six instrument scores — every instrument counted exactly once, with no instrument weighted more heavily than another. The result (0-100) maps to a label: 75-100 RISK-ON, 60-74 SLIGHTLY RISK-ON, 40-59 NEUTRAL, 25-39 SLIGHTLY RISK-OFF, 0-24 RISK-OFF.

3

Sub-scores — only where real data supports them

Gold Sentiment = Gold's own score. Crypto Sentiment = Bitcoin's own score. Equity Sentiment = average of Nasdaq + S&P 500. Risk Sentiment = average of all six tracked instruments. We deliberately omit a “USD Strength” sub-score: this site does not track a standalone USD/Dollar-Index instrument in the bias engine (EUR/USD and GBP/USD were removed as standalone tracked instruments to reduce backend load — see below), so there is no real data to honestly derive that number from. We show the omission explicitly on the homepage rather than inventing a figure.

4

Previous / Change — real day-over-day comparison

Once per day, the current score is written to a small history table. The homepage shows “Previous: X / Change: +/-Y” only once a prior day's real snapshot actually exists in that table — on the very first day this feature is live, there is nothing to compare against yet, and the homepage says so honestly (“first day tracking”) instead of showing a fabricated change figure.

Tap/click the gauge on the homepage to expand the sub-score breakdown — it's collapsed by default so the headline number isn't buried under a wall of figures.

What this methodology does NOT claim

  • No proprietary, battle-tested trading algorithm — the pipeline above is the real one, in full.
  • No historical-accuracy track record or win-rate statistics are published, because we don't fabricate ones we haven't actually measured.
  • No guaranteed predictions or buy/sell signals of any kind.
  • No claim that AI classification is fact-checked to journalistic standard by humans on every story before publish — human review exists as an ongoing correction mechanism, not a pre-publish gate on every item.
  • No fabricated Market Mood Index sub-score — "USD Strength" is explicitly omitted rather than invented, because no standalone USD instrument is tracked by the bias engine.
  • No fabricated day-over-day Mood Index history — "Previous/Change" only appears once a real stored snapshot exists to compare against.
  • The Gold page's "Bullish drivers" / "Bearish risks" are excerpts of our own already-published Gold articles' stored reasoning — not independently generated generic talking points.
  • The Economic Calendar shows only the real United States (USD) High-impact feed we already ingest (Fair Economy / ForexFactory) — no other countries/impact levels are fabricated to fill out the table.

See also our Trust & Sourcing page and our About page.