hirenyi/easyquant

EasyQuant is a platform for researching and testing stock trading strategies with the help of AI agents, built on top of an existing project called QuantGPT. It adds automatic backtesting on JoinQuant, a Chinese stock market simulation website, along with batch strategy submission and a library of already validated strategies. The base QuantGPT part works through an AI agent, meant to be Claude accessed through the Model Context Protocol, that designs trading factor formulas on its own, tests them, scores them, checks for overfitting, and can submit promising results to WorldQuant BRAIN, a professional quant research platform. The agent has access to eight tools for running backtests, scoring factors, diagnosing why a factor failed, running anti overfitting checks, validating expression syntax, and listing available math operators and stock universes. The EasyQuant extension adds browser automation, using Playwright, that logs into JoinQuant, submits strategy code, waits for the backtest to finish, and scrapes back the performance numbers. Strategies are submitted one at a time because free JoinQuant accounts only allow one backtest running at once. The repository includes 28 strategies that have already been tested this way, each with performance numbers like annual return, maximum drawdown, and Sharpe ratio, along with the original QuantGPT project's three factors that were formally submitted to WorldQuant BRAIN. To run it, a person needs Python 3.10 or newer, Node.js for the frontend, Playwright with Chromium installed, and a free JoinQuant account, since non paying JoinQuant accounts only support one active backtest. A DeepSeek API key is optional and only needed if using the AI agent to design new factors rather than just running the already validated strategies. This project is aimed at people doing quantitative trading research on Chinese stock markets who want AI assistance combined with automated backtesting. It is released under the MIT license.