Research & Insights

Research & Insights

Engineering notes from the GodFin stack: where the models fail, how the trading engine adapts, and what we are learning while building an AI-native operator for active markets.

Published March 26, 2026

The Prediction Bias Problem in LSTM Trading Models

Why even strong sequence models can drift into persistent optimism or pessimism, and how rolling z-score de-biasing restores cleaner signal thresholds.

Read article →
Published March 26, 2026

How GodFin Reads Market Sentiment: Reddit, News, and Google Trends as Trading Signals

Why sentiment analysis matters and how GodFin aggregates 5 independent data sources into a unified signal that catches moves before the chart confirms them.

Read article →
Published March 26, 2026

How GodFin Scans 12,000+ Stocks Every Morning Before Market Open

A two-tier universe scanner that filters 12,627 tradeable stocks down to the top 20 opportunities daily—combining fast momentum screening with deep LSTM and sentiment analysis.

Read article →