Backtest Engine#

A professional-grade portfolio backtesting framework built on PyArrow for high-performance columnar data processing. Designed for quantitative analysts, portfolio managers, and researchers who need realistic, reproducible simulations of portfolio strategies with institutional-level cost modeling.

  • Configurable via Excel or programmatic dict/YAML

  • Flexible data ingestion — reads CSV and Parquet market data with user-defined column names

  • Dates in ``YYYY-MM-DD`` format — rebalancing runs on the exact date listed in the portfolio file (not t+1), using that day’s prices

  • Realistic cost modeling — separate prices for execution, commission calculation, and mark-to-market valuation

  • Multi-variation testing — run parameter sweeps (commission, capital, cash reserve, date windows) loading data once

  • Batch testing — auto-discover and backtest every portfolio in a directory with a shared market data pool

  • Comprehensive reporting — Excel reports with CAGR, Sharpe, Sortino, Alpha, VaR, CVaR, Max Drawdown, annual returns, drawdown analysis, and portfolio weights

  • Interactive dashboard — cumulative returns vs. benchmark, composition charts, drawdown visualization, and returns distribution

Installation#

pip install kaxanuk-backtest-engine --extra-index-url https://license:{YOUR_LICENSE_KEY}@{SERVER}/simple/

Note

The full command with your license key is included ready to copy in the welcome email you receive after purchase.

You can acquire your license here.

Interested in collaborating on a custom project? Get in touch at software@kaxanuk.mx.

Check the official release on PyPI for the latest version.

What’s new in 0.58.x#

Headlines from the recent 0.58.x release line:

  • Breaking. Environment variable renamed from KNPC_API_KEY_KAXANUK to KNBE_API_KEY_KAXANUK. Update Config/.env after upgrading.

  • New. arrow_to_pandas_for_analysis() helper converts a pyarrow.Table returned by the backtester into a pandas.DataFrame with plot- and analysis-friendly dtypes (float64 / datetime64[ns]), bypassing the object-dtype surprises of raw decimal/date32 columns.

  • Fixed. Embedded benchmark charts no longer show a spurious vertical line at 1970-01-01; the issue was traced to a RangeIndex being coerced as nanoseconds-since-epoch.

  • Fixed. MultiPortfolioBacktester.generate_individual_reports now sets the benchmark DatetimeIndex from the canonical date column instead of a literal that did not exist, fixing the Max Drawdown Date and the annual-returns table of the benchmark.

  • Fixed. Clearer errors when portfolio_directory resolves to zero portfolios, and when execution_price (removed in v0.57.0) still appears in a batch config — both surface now instead of failing silently.

See the Changelog for the full changelog.

User guide

Discover the fundamentals of the Backtest Engine, learn how to configure your environment, and explore the essential features to get started quickly.

To user guide

API Reference

Dive into the comprehensive API documentation, including modules, components, execution models, and public interfaces.

To API reference guide

Methodology

Understand the assumptions and calculations behind the engine: price roles, execution model, cash reserve logic, daily valuation, and performance metrics.

To methodology guide

Release Notes

Browse the full history of changes, improvements, fixes, and new features across releases.

To release notes