Quant Mashup - Tommi Johnsen Sentiment Analysis Series Part 3: Three Ways the Sentiment Model Can Fail [Tommi Johnsen]Every day, financial news outlets publish thousands of articles about publicly traded companies. For investors, the obvious question is: does any of it actually matter? If a headline says a company just signed a major contract or passed a clinical trial, should you expect the stock to move the next(...) Can LLMs Beat FinBERT for Stock Sentiment Trading? [Tommi Johnsen]Part 2 of a series. Part 1 covered building the hybrid classifier and validating it against Claude as ground truth on 991 headlines. This post reports whether the sentiment signals actually predict stock returns. The Question The academic evidence is clear: investor sentiment predicts short-term(...) FinBERT Is Wrong 83% of the Time on Positive Headlines: an LLM is Here to Help [Tommi Johnsen]If you’ve ever plugged financial news into a sentiment model and used it to trade, you’ve probably noticed something: the signals are garbage. The model says “positive” all the time, your positions lose money, and you start wondering whether sentiment analysis is just astrology for quants.(...) Stock Sentiment Indicators in U.S. Equities: and the research that supports them [Tommi Johnsen]Academic research treats investor sentiment as a systematic component of beliefs or demand that is not justified by available fundamentals, and whose price impact is amplified when limits to arbitrage make it difficult for rational traders to offset mispricing (e.g., Shleifer and Vishny, 1997; Baker(...) Target-aware Financial Sentiment: Why Structure Beats Confidence with LLMs [Tommi Johnsen]Sentiment analysis in finance typically treats sentiment as a property of text as a whole, but what matters to investors is sentiment about specific entities. This paper investigates target-level sentiment attribution in financial news headlines, demonstrating that widely used sentiment models,(...) The Age of AI Attractor Markets: One Possible Trajectory [Tommi Johnsen]This essay explores one curious, if speculative, scenario for how markets might evolve. It’s not a forecast, just a framework worth examining. The core question: could modern markets be drifting from an exploratory regime with messy human disagreement, diverse models, and largely uncorrelated(...) AI is no longer an experimental “tool” in finance [Tommi Johnsen]But it is increasingly the way markets get read, sized up, and traded. If you work with stocks, you’ve probably felt it already. The advantage isn’t just speedy spreadsheets. The advantage is the ability to chew through messy, human language like headlines, filings, earnings-call transcripts,(...) What Investors Should Know About Common Sentiment Models: Tone Isn’t Attribution [Tommi Johnsen]Sentiment analysis has become a foundational tool in modern investing. From academic papers to hedge fund dashboards, models like FinBERT and RoBERTa variants are routinely used to classify news as positive or negative for stocks. However, there is a critical assumption hiding in plain sight. Thanks(...)