Data & badges

The entire Miracles Jar catalog — 178claims, each with its full assessment — is published as one open JSON file, and every claim has an embeddable verdict badge. The goal is simple: whichever side of an argument you're on, you can cite the same calibrated number from the same URL.

The dataset

One record per claim, generated at build time from the same loaders that render the site — the dataset can never disagree with the pages.

https://miraclesjar.com/dataset.json

Every number comes from the two-bar model: the Miracle Meter asks if the facts are true, could nature do this? and the Evidence bar asks did it happen as reported? The headline probability is their product — never 0 or 100 — and the pair sorts each claim into one of six tiers, from Gold standard to Disproven. Full definitions and thresholds are on the methodology page.

Field	Meaning
slug	Stable identifier; the claim lives at /claims/<slug>
title	Claim title
url	Canonical claim URL
type	What is claimed to have happened (healing, signs, relics, apparition, providence, eucharistic, baselines)
era	"modern" (2000+), "1900s", "earlier", or null when the date can't be inferred
tradition	Religious tradition, or null when genuinely unclear
mode	"law" = apparent violation of natural law; "providence" = extraordinary timing/arrangement
tier	gold · silver · bronze · explained · unproven · disproven (null = ungraded)
miracleMeter	1–99 — if the facts are true, could nature do this?
evidence	1–99 — did it happen as reported?
headline	1–99 — round(miracleMeter × evidence ÷ 100), the number shown on the claim page
confidence	low · medium · high — how much weight the estimate can bear
docScore	0–10 documentation depth
primaryRival	The natural hypothesis doing the most work (chance, expectancy, natural-course, misdiagnosis, fraud, misperception)
kernel	The minimal disputed facts the Evidence bar scores (present where defined)
disprovenKind	For Disproven entries: "confessed" or "refuted" (present where defined)
sourceCount	Number of sources in the claim's ledger
published / updated	ISO dates — first published; most recent rating revision on /changelog (null = never revised)

License

The dataset is released under Creative Commons Attribution 4.0 (CC BY 4.0). In plain language: use it for anything — articles, papers, apps, sermons, debunkings — including commercially. The only condition is credit: name Miracles Jar and link back to miraclesjar.com. The license is also embedded in the JSON itself.

Suggested citation:

Miracles Jar, "Miracles Jar Verdict Corpus", https://miraclesjar.com/dataset.json (CC BY 4.0).

For a single claim:
Miracles Jar assesses this claim at 63% (Gold standard) —
https://miraclesjar.com/claims/antonia-raco-2009

Verdict badges

Every claim has a self-contained SVG badge at /api/badge/<slug> — the Verdict Seal, the tier, and the headline probability. Drop it into any page with a plain <img>; it carries its own neutral background, so it reads on light and dark pages alike.

<a href="https://miraclesjar.com/claims/antonia-raco-2009">
  <img src="https://miraclesjar.com/api/badge/antonia-raco-2009"
       width="320" height="64"
       alt="Miracles Jar verdict badge" />
</a>

Badges update automatically when a rating is revised — no need to re-embed. Unknown slugs return 404.

Honest caveats

These are calibrated editorial judgments under the published rubric, not measurements. Each claim page carries the written reasoning and the evidence the number rests on — cite the number with its page.
Assessments are revised on the record as documentation improves: every change, with the old value, new value, and reason, is at /changelog, and the updated field tells you when a record last moved.
No value is ever 0 or 100 — certainty is unattainable in both directions, so everything is clamped to 1–99 by design.
The system audits itself in public: tier distribution, histograms, and anchor-compliance checks are recomputed from the data at every build on /calibration.