Performance Metrics

Ferric Alpha implements the core factor-analysis metrics expected by alphalens users while exposing Polars DataFrames at the Python boundary.

Information Coefficient

ic = fa.performance.factor_information_coefficient(factor_data)
mean_ic = fa.performance.mean_information_coefficient(factor_data)

IC is computed as cross-sectional Spearman rank correlation between factor values and forward returns.

Factor Portfolio

weights = fa.performance.factor_weights(factor_data)
returns = fa.performance.factor_returns(factor_data)
alpha_beta = fa.performance.factor_alpha_beta(factor_data)

Weights can be demeaned, group-adjusted, or equal-weighted. Factor returns use the same weight construction options.

Quantiles And Turnover

quantile_returns = fa.performance.mean_return_by_quantile(factor_data)
spread = fa.performance.compute_mean_returns_spread(quantile_returns, 5, 1)
turnover = fa.performance.quantile_turnover(factor_data, quantile=5, period=1)
rank_auto = fa.performance.factor_rank_autocorrelation(factor_data, period=1)

These APIs are designed for repeated factor research workflows: compare top-bottom spreads, inspect quantile stability, and monitor factor rank decay.

factor_rank_autocorrelation automatically uses a dense-matrix fast path when every date contains the same complete asset universe with finite factor values. Sparse or changing universes and invalid factor observations fall back to the general alignment path, preserving the same output contract.