StageAnalysis
Advanced stock market stage analysis platform with automated screening, AI-powered insights, and portfolio tracking.
The Challenge
Retail investors applying stage analysis methodology had to run their screening manually — downloading data, running calculations in spreadsheets, and manually tagging hundreds of stocks. The process took hours and was prone to inconsistency.
Automating a Manual Stock Analysis Workflow
Stage analysis — popularised by Stan Weinstein — classifies stocks into four stages based on price action and moving averages. Identifying Stage 2 breakouts early is the core trading edge. Doing it at scale requires consistent, automated calculation.
Data Pipeline
Market data is sourced via a daily batch job using yfinance and stored in PostgreSQL. A Python pipeline built on Pandas calculates 30-week moving averages, relative strength rankings, volume metrics, and Bollinger Band positions for each symbol.
The pipeline runs on AWS ECS on a nightly cron, with results materialised into pre-computed summary tables to keep dashboard queries fast.
Stage Classification
The stage classifier is a rule-based Python module with a TensorFlow-backed confidence scorer. Rules encode Weinstein's criteria explicitly (price above/below 30-week MA, MA slope, RS rank). The neural confidence scorer was trained on 10 years of historical price data labelled by a domain expert.
Frontend Dashboard
The Next.js frontend presents stocks in a sortable, filterable table with sparkline charts for the past 52 weeks. Users can save watchlists, set stage-change alerts via email, and view detailed technical breakdowns for individual stocks. All data is served from FastAPI with Redis caching for the most-queried endpoints.
Our Solution
We automated the entire pipeline: market data ingestion, technical indicator calculation, stage classification, and scoring run nightly across a universe of 500+ stocks. Investors log in to a ranked, filterable dashboard and receive alerts when stocks enter actionable stages.
Tech Stack
Results
- Full 525-stock universe screened and scored in under 5 minutes nightly
- Stage classification accuracy validated at 91% vs. manual expert review
- Dashboard API response times under 300ms with Redis caching
- Reduced analyst screening time from 4 hours to 10 minutes per session