A 3 GB Docker image that backtests stock algorithms at minute resolution

What it does

The Stock Analysis Engine is a data pipeline and backtesting framework for equities. It fetches minute-by-minute pricing, options, news, and fundamentals from IEX Cloud, Tradier, and FinViz; compresses and stores them in Redis; and runs algorithmic backtests against that history. Results and datasets can be archived to S3 for training deep neural networks. The whole stack deploys via docker-compose or Kubernetes with Helm.

The interesting bit

The project treats backtesting infrastructure as a data engineering problem first, trading problem second. It bundles Jupyter notebooks that compare DNN architectures trained to predict closing prices, effectively using the backtester as a feature generator for machine learning rather than just a strategy validator.

Key highlights

• Supports intraday, daily, and weekly fetches through CLI, API, or Kubernetes jobs
• Includes notebooks for running custom minute-resolution algorithms and replaying compressed trading histories from S3
• Backs up cached pricing data to AWS S3 via example Kubernetes jobs
• ~3.0 GB Docker image includes workers, backtester, Jupyter, Redis, and Minio
• Claims >150 million trading history rows generated from +5000 algorithms

Caveats

• Yahoo Finance API was disabled in January 2019 and is no longer available; IEX and Tradier require paid API tokens
• Docker Compose uses network_mode: "host", which breaks on Docker for Mac
• README appears to be from 2019; current data feed compatibility and maintenance status are unclear

Verdict

Worth a look if you need a batteries-included sandbox for combining algorithmic backtesting with TensorFlow/Keras experiments. Skip it if you want a lightweight, actively maintained trading framework or if you’re unwilling to pay for IEX/Tradier data access.