cpz-quant¶

cpz-quant is the open-source quantitative research engine from CPZ Lab: portfolio optimization, covariance estimation, risk measures, time-aware cross-validation, anti-overfitting strategy certification, and vectorised technical indicators for Python.

pip install cpz-quant

Everything is a pure function: data in, results out. No I/O, no hidden state, fully typed, Apache-2.0.

Why cpz-quant¶

20+ portfolio allocators sharing one input shape: mean-variance, HRP, HERC, NCO, Schur, Black-Litterman, entropy pooling, mean-CVaR, robust MVO, risk parity, alpha-risk-cost, QUBO/quantum-inspired, and more.
DataFrame-native. Polars and pandas DataFrames work everywhere returns go; pandas is never a required dependency.
Rust-accelerated. The bundled cpz_risk_rs crate (PyO3 + rayon) speeds up certification analytics, covariance, and indicators, with parity-tested NumPy fallbacks.
A certification layer. Probability of Backtest Overfitting (CSCV), Deflated Sharpe Ratio gates, and regime-conditional analytics turn "trust me, it backtests well" into a reproducible grade.
Exact convex programs when you need hard constraints: cardinality, turnover, gross exposure, and robust uncertainty sets via the optional cvxpy backend.
Battle-tested. This is the research core that powers the CPZAI systematic trading operating system in production.

Where to start¶

Quickstart: install and run your first optimization in 60 seconds.
Portfolio optimization: every allocator, covariance estimator, and cost model.
Convex backend: cardinality constraints and robust uncertainty sets.
Model selection: walk-forward and combinatorial purged cross-validation.
Certification: prove your backtest is not overfit.
Indicators: momentum, trend, volatility, volume, statistical.

Relationship to the CPZAI operating system¶

cpz-quant is the open research core. The proprietary cpz-ai SDK builds on it with live multi-broker execution, FIX connectivity, and market data access on the CPZAI operating system. Research is open; execution is a product.