FTB-C000630 / Formula component

Residual Tolerance

Residual Tolerance is the inclusive ratio-distance threshold used to classify trendline residuals.

Also known astolerance_bps

Intermediate Formula defined 1 linked topicEvidence-reviewed

Definitions

In plain terms

A tolerance of 75 basis points allows a bounded ratio error around the line.

Technical

epsilon_log = log(1 + tolerance_bps/10000); residual membership uses absolute log error <= epsilon_log.

Scope

Changing tolerance can change the winning consensus and should be versioned.

Formula

epsilon_log = log(1 + tolerance_bps / 10000)
LaTeX: \epsilon_{log}=\log(1+\frac{tolerance_{bps}}{10000})
SymbolMeaningUnit
tolerance_bpsresidual tolerancebasis points

Output unit: log-price

Examples

  • A reviewer computes Residual Tolerance on a small deterministic fixture and checks timing, units, and boundary behavior before publication.

Common misconceptions

  • Changing tolerance can change the winning consensus and should be versioned.

Concept relationships

Prerequisites

Related

Required by

Where this concept is used

Tutorials planned

These catalogued topics use this concept, but their complete build has not shipped yet.

  • D08-F01-A04 Important

Evidence and governance

  1. sklearn.linear_model.RANSACRegressor scikit-learn · first party technical publication

    Supports: preferred label, short definition, technical definition, formula

    Limits: This package enumerates every two-point hypothesis for small pivot sets instead of using random sampling.

  2. Linear Least Squares Regression NIST/SEMATECH · official standard

    Supports: preferred label, short definition, technical definition, formula

    Limits: A fitted line remains descriptive and is not evidence of a future price path.

  3. numpy.log NumPy · first party technical publication

    Supports: preferred label, short definition, technical definition, formula

    Limits: A log transform requires positive finite prices and does not by itself define a financial level.

  4. numpy.polyfit NumPy · first party technical publication

    Supports: preferred label, short definition, technical definition, formula

    Limits: The glossary freezes its own log-space OLS contract, diagnostics, and tie rules.

Reviewed by
fintech-builder-batch-011
Last reviewed
2026-08-09
Next review
2027-08-09
Record status
evidence reviewed

This record is evidence-reviewed and readable, but not yet promoted to published — it is served noindex,follow and excluded from the sitemap.

Written by

Fintech engineer building market-data and financial systems, and the author of every article, glossary record, and reference implementation on The Fintech Builder.