News Sentiment Index

The News Sentiment Index measures shifts in market news sentiment relative to its recent baseline, not raw positivity/negativity.

Why z-score? Financial news has a permanent bullish bias (~2:1 positive-to-negative articles). Raw averages always look positive, even when sentiment is deteriorating. The z-score measures how far current sentiment has deviated from its rolling 30-day average, surfacing actual mood shifts.

Calculation: Articles are classified as bullish (>0.3), bearish (<-0.3), or neutral via FinBERT. The bull-bear spread (% bullish minus % bearish) is decay-weighted, then z-scored against its rolling baseline. The z-score is scaled to -100/+100 via tanh for readability.

Reading the score: 0 = normal baseline. Positive = more bullish than usual. Negative = more bearish than usual. ±40+ = significant deviation.

Time decay: Half-life of 12h (24h window), 3d (7d), 7d (30d), 14d (90d). Recent articles carry more weight.

Sector scores: Each sector is z-scored against its own baseline, so a sector that's normally bullish doesn't permanently rank high — only if it's more bullish than its own norm.

Coverage: US-listed stocks (NYSE, NASDAQ, AMEX). ETFs excluded. Min 5 articles/sector, 3/stock.