Factor Exploration

Use this tab as a fast pre-check before running full experiments.

Select factors, universe, and horizon, then run analysis to review mean IC, Rank IC, IC IR, coverage, bucket spread, and diagnostics copy.

If results are promising, continue in Research Lab for full validation.

The i on each factor card opens consolidated notes; after a run, metric headers, chart titles, and diagnostic rows expose field-level tooltips.

Under every chart you will find a short rule-based interpretation (deterministic thresholds and shape checks—no LLM calls), covering rolling IC, decay, quantile structure, distribution shape, and multi-factor correlation; empty payloads still produce a clear, non-hallucinated note.

Use Single vs Compare beside the verdict: Single shows aggregate series from the multi-select response. Compare (best with two or more factors) adds a summary panel, per-factor metrics table, mean-IC bars, rolling-IC overlay, grouped decay by horizon, and grouped quantile bars by bucket. The table and mean-IC bars reuse the same analyze response; overlays issue one additional analyze call per selected factor (capped at five) on the client without changing backend research logic. When Compare is on with two or more factors, aggregate IC/decay/quantile/distribution charts are hidden to avoid mixing signals; the correlation heatmap still reflects the current multi-select response.