Skip to main content
Miracles Jar

Calibration

This page exists so readers can audit the system, not trust it: every number below is computed from the published grades at build time, and the anchor table runs the same checks that gate our releases.

Honest dating: all three rubric versions were developed during the site's initial calibration on June 11, 2026 — the revision tables below are the complete record of that process, public from day one, not a multi-year history. The record accrues from here.

Rubric

v4.2

Graded claims

252

Last updated

2026-07-06

Tier distribution

Where the catalog lands under the six tiers. A healthy distribution is bottom-heavy — most claims should not survive scrutiny.

TierClaimsShare
Gold52.0%
Silver124.8%
Bronze145.6%
Explained13654.0%
Unproven6425.4%
Disproven218.3%

Miracle Meter

If the facts are true, could nature do this? Claims per 10-point bucket.

1–9
77
10–19
48
20–29
34
30–39
17
40–49
11
50–59
13
60–69
14
70–79
13
80–89
19
90–99
6

Evidence

Did it happen as reported? Claims per 10-point bucket.

1–9
21
10–19
26
20–29
21
30–39
17
40–49
14
50–59
12
60–69
12
70–79
31
80–89
40
90–99
58

Biggest movers, rubric v1 → v2

The ten headlines the v2 migration moved most — the honest cost of changing the formula in public. Full record on the changelog.

Biggest movers, rubric v2 → v3

The headlines the v3 refinements moved most — the F-floor for refuted-but-unconfessed claims did the heavy lifting. Full record on the changelog.

Anchor compliance

Calibration anchors from the rubric — fixed points a re-grade is not allowed to drift past. The same rules run as a hard gate in CI.

Confessed hoaxes / exposed frauds: Evidence ≤ 5

Cottingley, Popoff, Trevignano — the record says it didn't happen, however high the meter.

PASS3/3

Resolved-naturally / boundary cases: Meter ≤ 5

Lentz, McMath, Ziadna, Lahaina — confidently documented, naturally explained.

PASS4/4

Headline = round(Meter × Evidence ÷ 100), clamped 1–99

Every headline is the product of the two bars — no hand-tuned exceptions.

PASS252/252

Gold tier ⇔ Meter ≥ 70 and Evidence ≥ 80

Gold is earned on both bars, in both directions — no entry holds it without the thresholds (Evidence bar raised 70→80 in rubric v3).

PASS252/252

All values in range

Meter, Evidence, headline within 1–99; docScore within 0–10.

PASS252/252

Thresholds, bars, and tiers are defined on the methodology page; every individual revision behind these aggregates is on the changelog.