Overview

Step-by-step guides for common factrix workflows:

• Copy-ready Plotly charts built from factrix producer frames — Plotting recipes
• Logging complete, auditable EvaluationResult records to MLflow — Integration adapters
• Raw data to a four-column factrix panel (sort order, frequency alignment, missing data) — Preparing data
• Picking the right metric for your research question — Choosing a metric
• Data structure dispatch (PANEL ↔ TIMESERIES) and the n_assets == 1 special path — Panel vs timeseries
• Reading EvaluationResult, MetricResult, and FDR results — Reading results
• False Discovery Rate (FDR) control across a factor batch — Multi-factor screening example
• Choosing evidence for allocation signals without turning factrix into a portfolio optimizer — Validating allocation signals
• Bounding peak memory when screening 100–1000+ factors (caller-side batched loop) — Large-scale evaluation and memory protection
• Slicing one metric by an attached label column (sector / regime / universe / ADV bucket) — Slice analysis
• Running every descriptive metric applicable to a cell — Standalone metrics
• Observability and DAG tracing (logger configuration and HTML rendering) — Observability
• Writing and registering custom metrics — Custom metrics