News velocity (v95) — model card

v1.0 · Tier 4claude-haiku-4-5Live Story-rate × sentiment composite.

Purpose

Catalysts and "storm" conditions don't move price linearly with story count — they move with story count × sentiment magnitude × novelty. v95 produces F08_news_velocity factor input + storm flags for the alert pipeline.

Inputs

• News headlines from /api/news + /api/news-aggregator (Yahoo/Finnhub/SEC EDGAR)
• Per-headline sentiment from /api/sentiment-analysis (Claude Haiku 4.5)
• Ticker-tagged stories (last 30 minutes for fast feed; 24h for full)

Outputs

{
  ticker:        'NVDA',
  storyCount30m: 7,
  storyCount24h: 42,
  sentimentMean: +0.34,                // -1..+1
  sentimentStd:  0.42,                 // disagreement signal
  novelty:       0.78,                 // vs trailing 7d baseline (1.0 = avg)
  storm:         true,                 // storm = ≥5σ above baseline
  composite:     0.62,                 // F08_news_velocity input
  topHeadlines:  [{ title, source, sentiment, ts }, ...]
}

Storm detection

storm = (storyCount30m / baseline_30m) ≥ 5σ AND |sentimentMean| ≥ 0.4. Flags fire alerts in the inbox + suppress new entries on the playbook (event-window disqualifier per playbook §"when NOT to act").

Limitations

Source

engine/v95-news-velocity.js, fed by /api/news-aggregator, /api/news-flow, /api/sentiment-analysis

Reviewed: 2026-04-27 · Next: 2026-07-27 · Per Bible §15.61