Who this is for — Traders blending fundamentals, technicals or quant with market «tone» — headlines, social, analyst notes. Useful as filter and confirmation, fragile alone.
Sentiment analysis quantifies attitude (positive, negative, neutral) toward assets, sectors or macro from textual and behavioural sources — news, earnings calls, Twitter/X, Reddit, AAII surveys, extreme positioning as euphoria/fear proxy.
In plain terms — Gauge whether the market is scared or euphoric — often a contrarian at extremes, moderately useful in the middle.
Sources and methods
| Source | Method | Limit |
|---|---|---|
| News / wire | Lag, duplicates, clickbait | |
| Social | Volume + polarity | Bots, manipulation |
| Surveys | % bullish/bearish | Small sample |
| Market | VIX, put/call, extreme funding | Not text, but «price sentiment» |
Sentiment combines with market regime and herding: persistent extremes often precede mean-reversion or final melt-up.
Operational use
- Confirmation: technical setup aligned with non-extreme sentiment
- Contrarian: record euphoria + price divergence → caution on longs
- Risk-off: panic + thin liquidity — watch gaps and slippage
- ML: feature in pipeline with lags and data leakage control
Common mistake — Trading real-time sentiment without news lag — price often already discounted the headline.
Example — Crypto fear & greed at 90 (extreme greed) + elevated BTC-PERP funding → reduce long size despite uptrend.
Summary card
- Role: context and filter, rarely sole trigger.
- Quality: source, timestamp, deduplication.
- Hub: AI & markets.