Skip to reversion research
Mean Reversion StrategyReference / Excursion / Evidence

Questions about this research library

No. Descriptive movement toward a reference omits execution, costs, sizing and adverse excursions. A usable rule needs an independently evaluated specification and an implementable instrument.

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.

Is mean reversion automatically profitable?

No. Descriptive movement toward a reference omits execution, costs, sizing and adverse excursions. A usable rule needs an independently evaluated specification and an implementable instrument.

Does a two-standard-deviation move imply a 95% event?

Not without an appropriate distribution, parameter and sampling model. Prices, fitted residuals and rolling z-scores are different objects. Dependence and estimation change the interpretation.

Can I use the worksheet half-life as a stop?

The worksheet performs conditional-expectation arithmetic from an assumed coefficient. It does not estimate a market process or prescribe a stop. Negative coefficients and phi at or above one are deliberately outside its lane.

Library

Is every crossing a successful recovery?

No. With frozen reference 100 and initial close 104, a later 95 crosses the reference but has absolute distance five rather than four. Under strict distance recovery it is not closer. Equal distance also fails that strict rule. A separate crossing label can be studied if it was defined beforehand, but cannot silently replace closeness after a favorable endpoint appears. The workbook retains both definitions and their first observation indices. No actual trade or return follows from either label, and reaching a reference at one time does not guarantee the terminal outcome remains favorable.

What does a missing hit mean?

It means the specified event did not appear among the inspected closes under its frozen operator and horizon. It does not establish that the process will never hit later. A complete supplied horizon with no hit differs from incomplete data; the lens rejects the latter instead of silently filling missing prices. Observation count is not elapsed calendar time without an actual venue/source clock. Later declared values can remain in the input while being excluded from the chosen horizon. Neither excluded later hits nor missing intervals may be used to extend a favorable window retrospectively.

Can the path lens estimate a mean-reversion probability?

No. It accepts one frozen reference and a short unverified positive-close path, then computes deterministic distances and endpoint indices. It fits no process, uncertainty interval, survival curve or stationarity test. Its presets are invented. Real first-passage evidence needs a complete eligible event population, dependence treatment and source timing. A probability grade cannot be recovered from one illustrated path or an old aggregate win percentage. The AR worksheet is also conditional arithmetic, not an estimator of how many real events will complete within a chosen deadline.

What makes a claim testable?

Name the actual process, permitted source, reference, scale, event filter, outcome ownership and frozen selection procedure. Include cases that fail, remain unresolved or are excluded. Then perform the full independently required financial research with native entries, exits and accounting. Actual release adds cloud identity, subscriber route and certification. Documentation can expose an unmet requirement without accusing a provider of misconduct or inventing a failure result. The workbench collects references and questions; selecting an evidence state does not manufacture an audit or admit a candidate.