6  The Nine-Gate Framework

A Diagnostic Protocol for Fat-Tail Signal Validation

Author

Jean-Marc Choufani

Published

January 1, 2026

Status: Draft in progress (paper5_gauntlet_framework.md)


NoteChapter in progress

This chapter is under active development. The framework is designed and the gates are defined; the full write-up is forthcoming. The working draft is available on request.

6.1 Overview

The preceding four chapters document specific methodological failures: look-ahead contamination, estimator inconsistency under mixture processes, and incomplete bias correction. A practitioner encountering a new signal needs a systematic protocol for checking which of these failures — and others — may be active simultaneously.

The Nine-Gate Framework is that protocol. Each gate is a binary pass/fail test. A signal must pass all nine gates before it can be described as validated under fat-tail conditions. The gates address:

Gate Threat addressed
Gate 1: Point-in-time discipline Look-ahead contamination (Chapter 4)
Gate 2: Prospective vs contaminated divergence Problem 2 bias (Chapter 4)
Gate 3: Source-of-tail pre-classification Hill estimator domain of validity (Chapter 3)
Gate 4: Distribution-free test statistics Mean inflation under fat tails (Chapter 1)
Gate 5: Multiple-testing correction Deflated Sharpe Ratio / PBO
Gate 6: Effective sample size Serial correlation via Newey-West (Chapter 4)
Gate 7: Regime robustness Single-regime artifacts
Gate 8: Mechanism plausibility Economic mechanism, not data mining
Gate 9: Out-of-sample accumulation Live forward validation

The framework operationalizes the epistemological argument running through this monograph: a result that cannot survive these nine gates is not a finding — it is a measurement artifact.


Full chapter forthcoming. Subscribe to the Substack for updates, or watch the GitHub repository for new commits.