Skip to reversion research
Mean Reversion StrategyReference / Excursion / Evidence

Sources, corrections and evidence gaps

The guide bibliography links original statsmodels and SciPy documentation. The chart is generated from the disclosed worksheet assumptions; it is not a downloaded price history. The legacy lab accepts a local price list and does not fetch quotes.

Original editorial revision: 9 October 2026. Expanded edition: 14 October 2026.

Educational diagnostics, not a signal service. A standardized price displacement is not a calibrated probability; invented examples and user-supplied observations are not verified market performance.

Source map

The guide bibliography links original statsmodels and SciPy documentation. The chart is generated from the disclosed worksheet assumptions; it is not a downloaded price history. The legacy lab accepts a local price list and does not fetch quotes.

Correction and operator gaps

A monitored correction or security mailbox, operator details and commercial disclosure have not been confirmed. Old contact strings are not proof that a working channel exists. This page records that gap instead of giving a fictional address. Versioned source files and the dated publication report retain the changes.

Statistical documentation supplies definitions

The source map links official autoregression, unit-root, stationarity and standard-score references. These explain their respective specifications, not the behavior of a proprietary trading model. A test result depends on process choice, deterministic terms, lags and sample; this edition does not fit those tests to genuine prices. The AR worksheet converts an assumed positive coefficient into expectation half-life. The new path lens measures supplied distances without estimating that coefficient. A source for a statistical mechanism should not be presented as independent evidence of a profitable entry rule or an executable exit.

Price identity comes before a recovery count

A genuine source needs instrument, adjustment, venue calendar, version and availability. Revised data may describe the past differently from the version available at a decision. Missing bars and zero-scale exclusions belong in the population record. A spread additionally needs both legs, alignment and a frozen ratio. Do not infer those objects from a CSV's order or modification time. The original lab accepts local positive closes and uses prior observations, but cannot authenticate the user's source. Its descriptive count is not automatically a complete historical opportunity ledger or an independent validation sample.

Every teaching declaration can be inspected

The excursion ledger includes step zero and every future declaration from all seven cases. A marker distinguishes inspected observations from later unused values in the bounded case. That completeness concerns the authored population only. Its artificial steps have no genuine elapsed-time interpretation. The compact packet keeps reference, band, horizon and exact calculated metrics together. No missing market bar, trade or customer record is invented to fill a table. Copies downloaded from the publication remain fictional even when a user independently reproduces every metric. Reproducibility of a teaching example does not make its source authentic.

Corrections should keep the prior scope visible

A corrected reference rule can change outcomes without changing prices. Preserve the original rule, successor and which labels each owns. A revised source can also change the event population, requiring a separate compatibility decision. Original article dates remain visible next to the expansion date. Public third-party documentation and terms can change after review; this page makes no unattended monitoring promise. TODO(Darren): supply lawful primary records and responsible publisher contact before any measured vendor claim is strengthened. Unknown historical cutoff, calibration and account accounting remain explicit rather than being repaired with plausible numbers.