Sentiment analysis uses natural-language processing (NLP) to score how positive or negative a piece of text — a headline, an earnings call, a flood of social posts — appears to be. The pitch: turn the crowd's mood into a tradeable signal. The reality is more nuanced.

What it can do

Sentiment can add context: a sudden spike in negative news flow around a ticker, or a shift in tone during a Fed statement, is real information a model can quantify faster than a human skimming headlines. As one input among many, it has value.

Where it fails

Three problems. It's noisy — sarcasm, nuance, and mixed messages fool scorers. It's laggy — by the time news is text, the fast money has often already moved. And it's gameable — social sentiment in particular is trivially manipulated by coordinated posting or bots.

By the time a mood is measurable in text, the market has usually already priced it. Sentiment is a rear-view mirror wearing a costume.

How to use it well

Treat sentiment as a secondary context layer, never a primary trigger — a reason to expect wider swings or to stand aside, not a standalone buy/sell. Price and structure lead; sentiment supports. See trading the economic calendar for how scheduled catalysts reshape conditions, and what moves SPY.