Research Methodology, Sources & Corrections

A claim should be traceable.

Research on this site is published under the InsomniCapital organisation byline. Each piece should state its question, input period, source, method and limits. We do not claim independent expert review, professional qualifications or a verified live trading record for these studies.

A publication date records when a study was released. A data cutoff records the latest observation used. A source-check date records when supporting material was inspected. These dates answer different questions: publishing a study in 2026 does not turn its 2024 observations into current market information.

Three kinds of evidence.

Historical observations

Identify the provider, units, observation dates and source vintage. Reference exchange rates are not executable quotes or intraday candles. Preserve the downloaded sample and explain transformations, missing dates and selection rules.

Hypothetical models

Label every assumed price, cost and rule. Show formulas and sensitivity to changed inputs. A calculated break-even threshold does not estimate the probability of achieving it.

Schedule calculations

State the session convention, named time zones, weekday filters and treatment of holidays. Future calendar projections are not observed liquidity, volatility or actual dealing hours.

What the checks cover.

Numerical examples are compared with independently calculated values. Tests exercise boundaries and invalid inputs as well as the usual example. Source citations are checked for the relevant definitions and data conventions. Charts identify their units, period and whether the values are observed or hypothetical; tables or downloadable data provide a way to inspect the underlying values.

Browser checks cover narrow and wide layouts, navigation, keyboard use and accessibility rules. These checks can catch arithmetic, content and interface problems. They cannot establish whether a strategy has a durable edge, whether a market relationship will persist or whether trading is suitable for a particular reader.

We keep research separate from service promotion. A relevant link may lead to the free Axi signals offer, and InsomniCapital may receive compensation for qualifying referrals. That commercial relationship is disclosed and does not make a research scenario an account offer or a verified performance record.

Reproducing a result.

1. Read the source period, quote direction and sample-selection rule before comparing numbers.
2. Use the accompanying CSV or JSON and preserve its units. For sourced observations, compare against the original provider and note any later revisions.
3. Apply the formula without intermediate display rounding. Record conventions such as sample versus population deviation, return interval and annualisation factor.
4. Check at least one result with a separate calculation. A result from the same implementation is not an independent confirmation.
5. Compare the limitations with the question you are actually asking. Historical reference-rate volatility is not a broker execution test, and a cost model is not a backtest.

Public studies expose their calculation method and downloads without requiring an account. The historical lab lets readers change windows using the same disclosed quote basis. Different providers, time boundaries or revised data can produce different results without implying that one is a current executable price.

Corrections should leave a visible record.

A substantive correction updates the affected prose, table, chart and download together. Its dated note should explain the error and whether it changes the conclusion. Layout repairs do not imply a fresh review of market facts. If a figure cannot be supported, it should be removed or clearly marked unresolved until it is checked.

The corrections section on each new study records its initial publication and any later changes. Source-data revisions should be distinguished from mistakes in our own calculations. A new data vintage should state what changed rather than silently replacing a historical sample under an unchanged description.

For a useful correction report, identify the page URL, the exact claim or number, the source and date you compared it with, and the calculation needed to reproduce the difference. Do not include account credentials or private financial records.

The current release contains a historical FX study, a hypothetical gold-cost study and a session-calendar analysis. There is no automatic daily news publication or promise of future coverage. New commentary requires fresh source checks and the same publication review before it is added.