Methodology
Research methodology and corrections
Our reports separate published evidence, our calculations and illustrative scenarios. Every claim should be read with its source, period and limitations.
Sources and scope
Read each linked report's publication date, cited sources, methods and limitations as supplied; a source citation is not independent verification.
How AI is used
We use current AI models to gather and aggregate source material and to help prepare each report. Findings may combine third-party evidence, our own calculations and illustrative scenarios, and each report says which is which. Original authors and data publishers are credited with direct links. Read more about BlackRidge.
Evidence hierarchy
A backtest, a forward window and a live track record are not interchangeable kinds of evidence. When reading any report on this site, check which kind of evidence a figure rests on before treating it as an expectation.
Costs and risk
Where a figure can be stated gross or net of fees, the report should be read with the fee structure it assumes, including management and performance fees, spreads and leverage. Tail losses can be sudden and large, and where an allocation is notionally funded the funded amount can be lost in full. Nothing in the research promises that a deposit caps losses.
Limitations and correction requests
Source selection, missing observations and updated evidence can all change what the data supports, and figures should be read with the period they cover. If you believe a report contains an error, write to us at the email published on this website and we will review the evidence and correct confirmed material errors.
Suggest a correctionCitation and reuse
Cite the original work for a result it established, and cite our report for our synthesis, explanation or an identified recalculation. Each report page has a Cite this report block with a reference, BibTeX and a stable link. We do not require a link, a particular anchor text or a link in return, and inclusion of a source does not imply that its authors reviewed or endorsed BlackRidge.
Resources for writers and researchersHow to read the evidence
First identify the sample, frequency and period behind a result. Reconstructed returns from published predictors are not verified live trading. A forward window in historical data is not proof of an investor's realised experience. The forward-validation report makes this distinction between reconstructed backtest, forward-window and post-publication returns explicit.
Then identify the quantity being measured. Covariance forecast error, realised volatility, expected shortfall and drawdown answer different risk questions. The covariance report's citation context states the portfolio sample and what its stress-risk finding does and does not establish. The loss-tail reports compare adverse outcomes with normal-model forecasts rather than treating a Sharpe ratio as a complete description of downside risk.
Finally separate an empirical result from an illustration. The fee calculator applies the same assumed gross return path to two fee structures. It isolates fee arithmetic under stated assumptions, excludes broker costs and taxes, and is not a return forecast or an assessment of a manager's skill.
Examples
Examples of the practice described above, including how evidence is separated from calculation:
Put the method in context.
Read a report with its sources, sample and limitations alongside the findings.
Next: Start here