numerai-experiment-design

What problem does it solve?

This Skill provides a structured workflow to design, run, and document Numerai experiments, enabling disciplined, iterative research across rounds.

Core Features & Use Cases

• Structured experiment planning: define baseline, feature_set, metrics, and decision rules for progression.
• Round-based evaluation: run multiple configs per round, synthesize results, and decide next steps.
• Reproducible experiments: maintain an experiment.md with decisions, results, and rationale to ensure traceability.

Quick Start

Create a new experiment directory under agents/experiments (one line per idea). Initialize baseline and initial configs, then run the training and evaluation loop with PYTHONPATH=numerai python3 -m agents.code.modeling --config <config> --output-dir <experiment_dir>. Update experiment.md after each round with results and decisions.