Research reference
Research validation
Keep sample design, rejected ideas and reproducible evidence together.
Research validationResearch validationResearch validationResearch validation
Validate in time and remove overlapping outcomes
Separate model fitting from later evaluation, including the less visible leakage created by forward-return labels and panel rows.
Count the experiments that did not make the chart
Treat windows, filters, weighting rules and implementation choices as research trials, not invisible adjustments around a selected winner.
A benchmark comparison is not the same as alpha
Separate active return, factor-explained return and residual performance, and make benchmark and funding conventions comparable.
What an inspectable equity research dossier should contain
Organize hypothesis, data provenance, decision rules, validation and unresolved limitations so a reviewer can trace a result back to its inputs.