Explicit methodology
How we score
A score that cannot be checked is not a score. Below is the complete description of the scale, the weights and the editorial rules — enough for anyone to recompute our results and show us where we are wrong.
The governing principle
Unresolved does not mean extraterrestrial. But it does not mean explained either.
Two errors dominate this field and they are mirror images. First: “we do not know what it was, so it must be something extraordinary.” Second: “most cases have an ordinary explanation, so every case has one.” Both are unlicensed logical leaps, and both produce the same result — people stop looking at the data.
This entire database exists to keep a third category alive: I do not know, and I know exactly what is missing before I could.
Eight evidence axes
Every case is scored 0–5 on eight independent axes. The axes measure how well documented a case is, not how anomalous it is. A solved case can score highly if it is exemplarily documented — as the Chilean Navy footage of 2014 does.
Witnesses
Qualifications and independence of the observers.
Radar / sensor
Independent instrumental measurement — radar, ELINT, sonar.
Optics / IR
Imaging record: photography, film, infrared.
Physical trace
Material, ground trace, measurable effect on the environment.
Multi-channel
How many mutually independent information channels exist.
Documentation
Quality of the timeline and the surviving case file.
Anomaly
Strength of unusual features AFTER obvious errors are removed.
Data today
Whether the case can be independently reconstructed now.
Weights
The axes are not equal. Instrumental evidence and present-day verifiability weigh more than a judged sense of strangeness:
R 1.35 · P 1.35 · M 1.25 · D 1.15 · O 1.10 · T 1.10 · S 1.00 · X 0.90
The X (anomaly) axis carries the lowest weight deliberately. It is the only axis requiring a subjective judgement from us, so it must not dominate the result. The D (data today) axis is weighted high, because a case that cannot be checked today is of less research value than a weaker but verifiable one.
Evidence classes
A class is derived from the scores, deliberately modelled on the French GEIPAN scale:
- A — evidential core. High score, instrumental evidence, many independent channels.
- B — strong documentation. Good score and at least one instrumental record.
- C — material gaps. A real case with incomplete material.
- D — testimony only. No evidence beyond an account.
Five source tiers
This is the most important element of the whole system. Every source on every case carries a tier. The colour beside a citation tells you what a given sentence rests on.
Primary document
Operational report, case file, gendarmerie protocol, lab result — produced close to the event.
Official statement
Institutional confirmation: ministry statement, authenticity release, agency determination.
Sworn testimony
Witness account given under legal weight — congressional testimony, sworn statement.
Later account / reconstruction
Interview, memoir or reconstruction produced years after the event. The layer most prone to growth.
Media claim
A claim circulating without documentary backing. Recorded because it shapes how a case is read.
State does not mean investigative
A state news agency dispatch is still T5 if no investigation stands behind it. A source’s tier is set by how it was produced, not by the prestige of its publisher. That is why the 1989 TASS report ranks below a French gendarmerie protocol from a rural property.
Editorial rules
- An authentic recording is not an identified object. The Pentagon confirmed three Navy recordings are authentic. It did not confirm what is on them. Those two sentences are never merged.
- Without range there is no speed. Any claim of extreme performance requires a known range to target. Where there was none, we mark it as a claim, not a measurement.
- A measurement is not the same as a software product. An airborne radar shows the output of a tracking filter, not a raw position. An IR camera shows a projection onto an observation plane from a moving platform. Before calling anything extraordinary, we rule out sensor physics.
- Without a chain of custody there is no material evidence. Analysis of a sample of unknown origin has no evidential value, however precise the analysis.
- A missing link is visible. Every case states how many of its sources have a link to the material. We do not hide gaps in provenance — that would be exactly what we criticise in others.
- A link to the archive is not a link to the document. Where we know which institution holds a record but cannot reach the record itself, we say so and point at the holding — but it does not count toward provenance. “We know where it is” and “here is the material” are different statements, and merging them is the easiest way to inflate an apparent evidence base.
- The sceptical hypothesis is stated as carefully as any other. If a prosaic explanation has a weak point, we describe it. Scepticism is not exempt from the burden of proof.
What this database does not do
It does not assign probabilities to the extraterrestrial hypothesis. We have no model that would license one, and a number produced without a model is worse than no number — because it looks like knowledge.
It also does not aggregate social media material. A screenshot from a messaging app destroys the data an analysis needs. Chile’s SEFAA requires original material, and that is the right standard.
Errors
This database certainly contains errors. Its complete revision history is public — every score correction, every change of source tier and every factual fix is visible together with its reasoning. Provenance applies to us as well.