How we built a Wall Street Sentiment analytics dataset for equity selection

How can you systematically replicate the reasoning of a seasoned financial analyst using AI and unstructured data?

In this whitepaper, we present backtested results from applying sentiment analytics to U.S. mid- and large-cap equities (June 2020 – June 2025).

Key highlights

• ~5% annualized returns from weekly sentiment-driven signals
• 0.90 information ratio
• 1-week effective holding period
• Controlled drawdowns (max ~6.5%)
• Consistent performance across daily, weekly, and monthly horizons

These results show that textual sentiment signals aren’t just promising on paper—they can enhance systematic equity selection.

How to create alpha signals your competitors don’t have

Wall Street Sentiment is just one application. The bigger opportunity lies in building custom datasets. With Bigdata Search, you can:

• Generate proprietary signals from billions of financial documents
• Automate complex analyst-style reasoning
• Build unique datasets tailored to your team, unavailable off the shelf

Owning your own data pipeline creates durable alpha others can’t replicate.

Get our playbook on replicating expert analyst reasoning with AI to sharpen your equity strategies.