Market Sentiment

Assets

Configured public universe: Gold, Silver, GDX, Bitcoin. Assets without a verified observation show as Unavailable — never as fabricated Neutral.

Data sources

• Official X API search when configured and enabled (bearer token; no scraping).
• Optional Reddit OAuth collection for related intelligence feeds.
• Mock/sandbox provider for development when live providers are offline.
• Collection runs on a scheduled cadence (~60 minutes), never on visitor page load.

Sentiment score

Classified posts receive lexicon labels mapped to a 0–100 scale (strongly bearish 10 → strongly bullish 90; neutral 50). Weighted average uses engagement/kind weights and classifier confidence. Uncertain and irrelevant posts are excluded — they are not forced to Neutral.

Status thresholds (centralized): ≥80 Strongly Bullish · ≥62 Bullish · ≥43 Neutral · ≥25 Bearish · else Strongly Bearish. Missing score → Unavailable.

Participation, crowding, divergence

• Participation / volume condition uses volume z-score versus recent history (not identical to raw count).
• Crowding combines extreme directional share, author concentration, and hot volume — separate from score.
• Divergence compares stored price history to sentiment bias with documented lookbacks; insufficient history is explicit.
• Percentiles require deeper history before public display treats them as established.

Freshness and last-known-good

Observations are append-only. Failed refreshes retain the last verified snapshot. Age-based freshness bands (Current / Delayed / Stale) do not reset a directional score to Neutral or zero.

Limitations

• English-oriented financial phrase classification; unsupported language is not silently Neutral.
• Bots, sarcasm, and symbol collisions (e.g. Barrick GOLD vs commodity gold) remain imperfect.
• Cross-asset scores are not perfectly comparable when source universes differ.
• Excluded from the Grey Rabbit composite score until validation completes.