pybroker

Why we included this project

PyBroker is a Python framework for algorithmic trading where strategies are driven by machine learning models rather than hand-written rules. Its backtesting engine is built on NumPy and accelerated with Numba, so sweeping across many configurations stays fast. A walk-forward analysis mode retrains models on rolling windows the same way they would be used in live trading, giving a fairer read on performance than a single hindsight backtest. Data can come from Yahoo Finance, Alpaca, AKShare, or your own provider, and the framework caches downloaded data, indicators, and trained models so iterative experimentation does not slow down. Traders of stocks or crypto who want one open-source tool that covers training, backtesting, and evaluation instead of wiring several libraries together should find this a good fit.