Methodology
How M5 turns a hypothesis into published, reproducible research.
Research Process
- Hypothesis — we start from a testable, falsifiable claim about market behaviour, framed so the data can prove it wrong.
- Data — we assemble the relevant market data with disciplined handling, documenting sources and treatment before any result is measured.
- Test — we evaluate the hypothesis against that data with transparent, pre-specified procedures, separating exploration from confirmation.
- Reproduce — we re-run the work from disclosed artifacts to confirm the result holds and is not an artifact of a single run.
- Disclose — work that clears the bar is published as abstracts and preprints with supporting code, disclosing the edge selectively and labelling honest status.
Reproducibility Stance
A result we cannot re-run is a result we do not claim. Wherever we report a finding, the goal is that it can be reproduced from the artifacts we disclose: methodology is documented before claims are made, and supporting code accompanies the papers it backs. Reproducibility is the default we hold our own work to, not a courtesy we extend after the fact.
Metrics-Disclosure Policy
Performance figures appear only alongside a linked, reproducible artifact and a clear disclaimer — never as a number standing on its own. Where a figure is illustrative rather than a live track record, it is labelled as such so the reader is never invited to mistake a demonstration for realised performance. The canonical wording we attach to any illustrative figure is: Illustrative backtest · see linked artifact for full specification. Until a figure can be tied to the archive and re-run from its disclosed artifacts, it is not published as a metric at all.