6 The Nine-Gate Framework
A Diagnostic Protocol for Fat-Tail Signal Validation
Status: Draft in progress (
paper5_gauntlet_framework.md)
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.