A standard for better research

Abstract

HCL-STD-001 defines how H Capital conducts, documents, evaluates, reviews, publishes, corrects and maintains research. Its purpose is to make research traceable, evidence-based, proportionate to uncertainty, reproducible where reasonably possible, and clearly separated from promotion.

It applies to research papers, frameworks, methodologies, experiments, datasets, research notes (published on the website as Quant Notes), failure reports, reviews, decision records, tool pages, and any public claim materially presented as the result of H Capital research.

Research Question

How should research at H Capital be conducted, documented, reviewed, published, corrected and maintained?

Hypothesis

Not applicable. This is a research governance standard, not an empirical study. Where a study does test a hypothesis, the standard requires it to be testable, falsifiable, written before the primary analysis, stated with a falsification condition, and preserved unchanged in the final document even when it is rejected.

Data

No empirical dataset applies to this governance document. The standard is the document itself, and the downloadable PDF is its authoritative text.

Methodology

Every controlled document has a permanent identifier of the form HCL-[CLASS]-[NUMBER] (standard, framework, methodology, research paper, research note, experiment report, dataset record, failure report, review, decision record). Governance is tiered: an internal research log, a research note or experiment (which may be published with a clear preliminary label, and includes Quant Notes), and a full research paper.

Each document carries a status (Concept, Proposed, In progress, Internal review, Published, Live test, Rejected, Under revision, Corrected, Superseded, Withdrawn, Archived). A study that tests a hypothesis also carries a verdict, kept separate from its status: Supported, Not supported, Inconclusive or Requires forward testing. A proposed study has no verdict.

The lifecycle runs from question through research plan, analysis, adversarial review and publication to correction or closure, with four stage gates. Confirmatory and exploratory analysis are labelled, every material change to the analysis is logged, and strategy research keeps an append-only variant register. Claims are classified (fact, measurement, observation, interpretation, inference and others) and sources are graded from A to D.

AI assistance must be disclosed, and a deciding number must come from a measured file, not from a conversation. Each publication gets a confidence classification (a six-dimension rubric) and a reproducibility classification, and states the review classifications actually used. Published work is corrected visibly and never silently replaced: a revised document gets a new version number, and the previous version stays available, marked as superseded.

The standard also sets rules for the website library (only categories with entries are shown, and placeholder entries are prohibited), for Quant Notes, and for tool pages, which are commercial pages, not research: they must not state win rates or returns and must not present backtests as evidence of future performance.

Results

The output is a published research governance standard, currently version 1.2 (30 September 2026). It supersedes version 1.1 (29 September 2026) and version 1.0 (1 August 2026). Version 1.1 named Tommy H as author and approval authority, aligned the statuses, verdicts and document classes with the public library, and added the variant register, the single-author independence rule, Quant Notes and the tool-page rules. Version 1.2 is a formatting correction only; no content changed. It is not evidence of a trading edge.

Robustness

The standard was produced with AI-assisted adversarial review and author self-review. Independent external peer review was not conducted. H Capital is currently run by one person, who is both author and approval authority; the standard requires that limitation to be disclosed on every Tier 3 publication. Its own classifications are confidence: High, reproducibility: not applicable. It must be reviewed at least every twelve months, and the next review is due 1 August 2027.

Conclusion

This standard provides the foundation for documenting, reviewing, publishing and correcting research at H Capital. Empirical verdict: not applicable.