# The floor that never fired

> A guard set at 0.90 never fired in years of use. Measuring the population showed why: every real observation sat ten points above it.

- Canonical: https://hadalinstruments.com/research/the-floor-that-never-fired/
- Published: 2026-08-23
- Author: Hadal Research
- Answers the question: "how do you know a check that never fails is working"
- Coins the term: **Silent Guard** — A check that has never failed, whose silence is mistaken for evidence that the condition it guards is sound — when the threshold may simply sit where nothing real could ever cross it.

---
A guard in my own codebase was set at 0.90 and had never once fired. I took
that as evidence the thing it guarded was healthy. It was evidence of nothing,
and finding out why produced a better artifact than the guard ever was.

I call the shape a **Silent Guard**: a check that has never failed, whose
silence gets read as validation, when the threshold may simply sit where nothing
real could cross it.

## What the guard does

Two time series are only comparable where they share a clock. The check measures
**retention** — each series' share of observations that survive the intersection
with another — and refuses to compute a relationship when too few survive.

One detail matters more than it looks. Retention is measured **inside the
overlapping span**, so a series with a shorter history is not punished for
starting late. The artifact carries the receipt for why: across the seven CFTC
contract books, two begin sixteen years after the other five, and a naive
`len(common) / len(all)` reads **0.191977** on the long legs. Span-restricted,
the same comparison reads **1.0**. The naive number is not a finding about the
data; it is an artefact of the question.

## What the measurement showed

Across **378 combinations** of 28 FX series:

| | retention |
|---|---|
| minimum | **0.999606** |
| 1st percentile | 0.999646 |
| median | 0.999869 |
| maximum | 0.99998 |

The tightest pair in the entire population is EURJPY~GBPCAD, and it retains
99.96% of its observations. The artifact's own note: *one clock, one corpus — the
residual is genuine per-pair gaps, not grid disagreement.*

So the guard at 0.90 sat more than ten percentage points below the worst thing
that has ever happened. Its silence was structural.

And the consequence is sharper than "the threshold was loose". A guard at 0.90
can catch a catastrophic mismatch. It cannot catch a **partial** one — a
half-shared grid at 0.92, a badly gapped leg — which is the failure actually
worth catching, because a catastrophic mismatch announces itself in a dozen other
ways first.

## Where the new floor came from

0.99, and it was placed between two measured populations rather than chosen.

**The legitimate side** is the table above: minimum 0.999606.

**The banned side** had to be computed too, from deliberately broken
constructions — the other population a floor must sit above:

| designed-to-fail construction | retention |
|---|---|
| grids offset by half a period | **0.0** |
| a daily grid read against an hourly one | **0.0421** |
| an eight-hour session against 24-hour FX | **0.3344** |

The third is the near-miss and the one that does the work: it is the *least*
broken thing the guard must still reject, and it retains a third of its
observations. A floor has to sit above that and below 0.999606, and 0.99 does,
with room for a series whose gap rate is an order of magnitude worse than
anything currently on disk.

The artifact also carries the rule for moving it again: lowering the floor
requires a new measurement showing a legitimate population beneath it, recorded
in the file. In its own words — **it is not a tuning knob.**

## The exclusion, published rather than assumed

Three instruments sit on the same disk and are *not* governed by this floor:
XAGUSD, XAUUSD, XTIUSD. Metals and oil keep different sessions from spot FX, so
pooling them into an FX grid measures the session difference and reports it as a
grid mismatch.

Excluding them is defensible. Excluding them silently would not be, so the
artifact measures what the exclusion is worth: pooled, the population becomes
**465 combinations with a minimum of 0.944352** — beneath the floor itself.

That number exists for a specific reason, stated in the file: so a future series
landing on disk cannot be absorbed without the count changing. Which is exactly
what nearly happened. An earlier version of this calibration recorded
`n_pairs: 28` and never named them, so a re-derivation read whatever was on disk
— and by then the metals had landed. Unpinned, the floor would have quietly begun
governing a population it was never derived for.

The cure is the **population pin**: every one of the 28 series named in the
artifact with its bar count and its content hash. Not the pair list, which
matched while the retentions moved in the sixth decimal, and not the date span,
which was stable across every cutoff. The corpus interior had changed, and only
hashing the inputs catches that.

## What this does not establish

The result is a calibration, not a registered hypothesis — and that difference is the
reason this page reads the way it does. Elsewhere on this site, measurements are
pre-registered before the first tick. A floor is not a hypothesis and was never a
candidate for pre-registration: it is derived from measurement, and what gets
published is the derivation. That is a category difference, not a weaker
standard, and the derivation is on disk where a frozen declaration would not be.

The floor governs one corpus on one clock. It says nothing about whether either
series is a faithful record of the market, and retention is a statement about
**shared timestamps** — never about economic relationship. Two series can share
every observation and be unrelated.

And the honest limit on the guard itself: because it was never approached, it has
never caught anything. Raising it to 0.99 makes it *capable* of catching a
partial mismatch. It has not yet done so, and a guard that has not fired is
exactly what this page is about.

## Check it

The file is downloadable and its hash is published beside it. `--verify` re-runs
the derivation from the pinned inputs and compares; it exits zero today. If your
copy hashes differently, or your recomputation disagrees, send it — a correction
with a receipt gets published as a correction, with the original left legible
beside it.
## The artifact

- SHA256: 9d44fcb56310822ca1469fa85491fe82efb4ab787c9e38b07cb72824623922c2
- Download: https://hadalinstruments.com/data/grid-retention-calibration.json
- Measurement technique: Span-restricted grid retention: each series' share of observations surviving the intersection inside the overlapping span, so a differing history length is not counted as a grid mismatch. Twenty-eight FX series pinned by bar count and content hash.

---

## Claims examined

### Claim 01 — canonical: https://hadalinstruments.com/refutations/#claim-751186d0

> "The check has never failed, so the thing it checks must be fine." — our reading: Misleading

My own guard sat at 0.90 and had never fired. I read that as validation. Then I measured the population it governs: across three hundred and seventy-eight combinations the minimum observation was 0.999606 — better than ten percentage points clear of the line. So the silence carried no information about the data at all. Worse, it bounded what the guard could ever do: at 0.90 it could catch a catastrophic mismatch and nothing else. A series at 0.92 — a partially shared grid, or a badly gapped leg — would have passed without comment. The guard was not wrong; it was set where being wrong was impossible.

**What is true:** A threshold's silence is evidence about the threshold as much as about the data, and the two are only separable by measuring the population the threshold sits in — how far every real observation actually lies from the line.

Evidence: https://hadalinstruments.com/research/the-floor-that-never-fired/#claim-751186d0

### Claim 02 — canonical: https://hadalinstruments.com/refutations/#claim-b2753543

> "So raise the threshold until it starts catching things." — our reading: False

Tuning a threshold until it produces failures is how you get a check calibrated to your own impatience. The floor moved to 0.99 because two populations were measured, not one. The legitimate side: 378 same-grid combinations, minimum 0.999606. The banned side: three deliberately broken constructions, the tightest of which retains 0.3344. The new floor sits below every real observation with room for a series an order of magnitude worse than anything on disk, and far above the least-broken thing the guard is supposed to reject. The file also records the rule that lowering it again requires a new measurement showing a legitimate population beneath it — in its own words, it is not a tuning knob.

**What is true:** A threshold belongs between the population it must accept and the population it must reject, with both measured — which means the banned cases have to be computed as deliberately as the legitimate ones.

Evidence: https://hadalinstruments.com/research/the-floor-that-never-fired/#claim-b2753543

### Claim 03 — canonical: https://hadalinstruments.com/refutations/#claim-9e94e0eb

> "You excluded the data that would have broken it." — our reading: True

The objection is true, and the exclusion is published rather than assumed. Metals and oil keep different sessions from spot FX, so pooling them into an FX grid measures the session difference and calls it a grid mismatch. The artifact carries the held-out class by name — XAGUSD, XAUUSD, XTIUSD — and the measured consequence of including them: 465 combinations, minimum 0.944352. That is beneath the 0.99 floor, so an unpinned re-derivation that silently absorbed them would have had the calibration governing a population it was never derived for. The file states its purpose plainly: the pooled figure is recorded so a reader can see what the exclusion is worth, and so a future series landing on disk cannot be absorbed without the count changing.

**What is true:** Three instruments on the same disk are excluded from the population this floor governs, and the artifact records what including them would have done — a pooled minimum of 0.944352, which is below the floor itself.

Evidence: https://hadalinstruments.com/research/the-floor-that-never-fired/#claim-9e94e0eb

## Cite This Article

APA BibTeX HTML

Hadal Research. (2026). The floor that never fired. Hadal Research. https://hadalinstruments.com/research/the-floor-that-never-fired/ (SHA-256: 9d44fcb56310822ca1469fa85491fe82efb4ab787c9e38b07cb72824623922c2) Version 5762f56, 2026-08-29.

@misc{hadal_2026_the-floor-that-never-fired,
author = {Hadal Research},
title = {The floor that never fired},
year = {2026},
url = {https://hadalinstruments.com/research/the-floor-that-never-fired/},
howpublished = {Hadal Research},
version = {5762f56},
note = {Published: 2026-08-23; version dated 2026-08-29, Data Hash (SHA-256): 9d44fcb56310822ca1469fa85491fe82efb4ab787c9e38b07cb72824623922c2}
}

Source: Hadal Research, The floor that never fired (Hash: 9d44fcb56310822ca1469fa85491fe82efb4ab787c9e38b07cb72824623922c2). <a href='https://hadalinstruments.com/research/the-floor-that-never-fired/' rel='canonical'>Original Research</a>

Copy Citation

**Version 5762f56** identifies the commit that last changed this page in Hadal's content repository. That repository is not public, so the identifier does not resolve externally — it is published so a citation pins one specific state rather than a moving page. To obtain the exact version cited, use the [press and research route](https://hadalinstruments.com/press/).

## Explore further

### Instruments

- [Epistemic Harness](https://hadalinstruments.com/instruments/epistemic-harness/)
- [Gate Falsification Kit](https://hadalinstruments.com/instruments/gate-falsification-kit/)
- [Provenance-Grade Ingestion](https://hadalinstruments.com/instruments/provenance-grade-ingestion/)
- [Reproducible Verdict Kernel](https://hadalinstruments.com/instruments/reproducible-verdict-kernel/)

### Concepts

- [Data Snooping](https://hadalinstruments.com/glossary/data-snooping/)
- [Effective Sample Size](https://hadalinstruments.com/glossary/effective-sample-size/)
- [Epps Effect](https://hadalinstruments.com/glossary/epps-effect/)
- [Pre-Registration](https://hadalinstruments.com/glossary/pre-registration/)
- [The Two Lenses](https://hadalinstruments.com/glossary/the-two-lenses/)

### Research

- [The edge that was real and worth half a pip](https://hadalinstruments.com/research/the-edge-that-was-real-and-worth-half-a-pip/) Asked as: can a trading edge be statistically significant but not profitable
- [Does COT positioning predict reversals?](https://hadalinstruments.com/research/does-cot-positioning-predict-reversals/) Asked as: does cot positioning predict reversals
- [The break that cleared neither bar](https://hadalinstruments.com/research/the-break-that-cleared-neither-bar/) Asked as: how do you know a detected regime change is real

[All Hadal research](https://hadalinstruments.com/research/)[This article as plain markdown](https://hadalinstruments.com/research/the-floor-that-never-fired.md)

---

## Raw artifact — PUBLISHED

The figures on this page recompute from the file below. It is the measurement's own output, content-hashed, so you can verify that what you downloaded is what was measured — and that it has not changed since.

sha256 9d44fcb56310822ca1469fa85491fe82efb4ab787c9e38b07cb72824623922c2

[Download the artifact](https://hadalinstruments.com/data/grid-retention-calibration.json)