Skip to content

FINSABER 2.0.1: FINSABER-2 Upgrade

Latest
Latest

Choose a tag to compare

View all tags
@waylonli waylonli released this 24 May 01:11
· 41 commits to main since this release
v2.0.1
4d3be1e

FINSABER has been upgraded to FINSABER-2.

This release moves the active main branch to the FINSABER-2 framework and publishes the reusable backtesting package workflow around explicit data, execution, and cost assumptions.

Highlights

  • Package-oriented backtesting framework available from PyPI as finsaber==2.0.1.
  • New parquet dataset support through FinsaberParquetDataset and the public dataset at FINSABER-V2-Data.
  • Explicit execution timing, including next_open and same_close, to make look-ahead assumptions visible.
  • Split-adjusted OHLC handling for realistic price simulation across stock splits.
  • Slippage, liquidity caps, commissions, and LLM cost accounting integrated into the backtesting cost model.
  • Structured result artifacts, including run config, manifest, summary CSVs, metrics, equity curves, trades, orders, rejected orders, and LLM costs.
  • Professional documentation site at waylonli.github.io/FINSABER.

Install

pip install finsaber==2.0.1

For local development:

git clone https://github.com/waylonli/FINSABER
cd FINSABER
conda create -n trading python=3.10 -y
conda activate trading
pip install -e ".[dev,research]"

Branch Notes

  • Use main for FINSABER-2 package development and new backtesting workflows.
  • Use reproduce if you want to reproduce the original FINSABER-1 experimental results with the v1.0 framework and legacy aggregated dataset.

Documentation

Read the documentation for beginner-friendly usage guides, framework architecture, execution settings, data loader contracts, strategy examples, and result artifact explanations:

https://waylonli.github.io/FINSABER/

Assets 2
Loading