Method
Source, confidence and conflict
This page explains how the files are built: when a figure gets published, what the confidence level says, and why conflicting values are not hidden.
Three confidence levels
Every value carries one of three states. The state does not say the value is correct; it describes the quality of its source.
The manufacturer's or the institution's own statement. A link is given; if the link dies, an archive copy replaces it and the date is recorded. The official badge shows who the statement came from; it does not show that the information has been independently verified.
Secondary source: news report, magazine, trade show or conference note. Unverified, but on record and not hidden.
Independent assessment with no official confirmation. Its badge is red, because this is an inference rather than a statement.
The three states differ by pattern as well as colour: solid border, dashed border, red border with a soft fill. They stay readable in colour blindness and in black and white print. The same three states look identical in the table, the map legend, the 3D labels and the timeline.
What a value carries
A figure without a confidence level never reaches the interface. There is no exception: the field stays empty rather than being filled with an invented value.
- Value
- The figure itself and its operator, if any. "> 280" is not the same value as "280"; when the source gives a bound, the bound is recorded.
- Confidence
- Official, press or estimate. It describes the quality of the source.
- Source
- The human-readable name of the body or publication that released the value, with a link where one exists.
- Document
- When a value is read from a document, the version of that document and the date we accessed it. The same address can be updated silently; without a version name, the claim that “this figure was written here” cannot be checked a year later.
- Digest and archive
- The SHA-256 digest of the document we read, and the address of a third-party archived copy where one exists. We do not republish the document from our own server: manufacturer cards carry fields we do not publish and carry the manufacturer’s branding, so serving a mirror would mean publishing them ourselves. The digest leaves a check without crossing that line: anyone who finds the same file computes the digest themselves.
- Verified on
- When the record was last checked. Six months from now, this field answers the question of where a figure came from.
The unit lives inside the field name rather than in a field of its own. Confusing metres with millimetres is therefore impossible.
Where there is no data, a dash is shown and screen readers hear "no data". Missing data is not a fourth confidence state; it says the record is empty there.
Why we do not hide conflicts
Open-source figures diverge from each other. Picking one and deleting the rest means making an editorial decision the reader cannot see. Instead they all sit stacked in the same cell, each carrying its own source.
Where only one figure fits, such as the scaled silhouette and the 3D model, the most reliable value is used: official, then press, then estimate. That choice affects the drawing, not the record — the table keeps showing all of them.
Four kinds of divergence
More than one value in the same field is not a contradiction on its own. Which case applies is computed from the operators and the scope fields, not flagged by hand. There are four outcomes.
- contradiction
- The intervals do not overlap: no number can satisfy both. Either one source is wrong, or the two are in fact talking about different things. This is rare, which is why the row is marked.
- different statements
- The intervals overlap, so both can be true at once. Lower bounds work this way: “> 280” and “> 500” do not contradict each other, the second already satisfies the first. This is a normal state, not a warning.
- different scope
- The values are not measuring the same thing: they were obtained by different methods, belong to different variants, or were stated on different dates. No interval arithmetic is done here, because the values are not comparable. Which field differs is written under the value.
- not comparable
- One of the comparability fields is filled on one side only. We do not know whether the values measure the same thing, so no interval arithmetic is done. Not knowing is not a contradiction: declaring one because a field is missing would be the opposite of the method this page describes.
Comparability is decided by two fields: how the value was obtained (declared, measured in a test, third-party measurement, inference) and which variant it belongs to. The date of the statement is shown under every value but does not enter the decision: two statements of the same scope can contradict each other on different dates, and that contradiction must stay visible. When these fields are unknown they are left empty and the value reads “scope not stated”. They are never filled in by guesswork.
The three rows below are HYPOTHETICAL. They belong to no system and are not taken from the files; they only show how the calculation works.
- “< 6 m” and “> 9 m” — contradiction. No length is both under 6 and over 9.
- “> 6 m” and “> 9 m” — different statements. Any body longer than 9 metres satisfies both.
- “9 m” (third-party measurement) and “> 6 m” (declared) — different scope. One measured the object, the other stated a figure.
- “9 m” (third-party measurement) and “7 m” (scope not stated) — not comparable. We do not know how the second figure was obtained.
The official badge shows who the statement came from; it does not show that the information has been independently verified. A figure being official does not mean we measured it, or that anyone else confirmed it.
What we do not publish
Some material is findable in open sources and still stays out of these files. The limit is editorial rather than technical.
- No target language. A range ring is drawn; no target city or country is labelled or highlighted. The terms used are "range envelope" and "reference point".
- No list of the target classes a system was developed against.
- No technical detail on production, guidance internals, warhead design or accuracy optimisation. The limit is this: dimensions, mass, programme timeline and published performance claims.
- No combat or explosion imagery.
This rule is not abstract. The primary source for the TAYFUN file is ROKETSAN’s product card, and there are two fields on that card we read and deliberately left out:
- Target type. The card lists the classes of target the system was developed against. That list does not enter the file: we draw the range envelope, but we do not build target language around it.
- Warhead type. The card states the type of warhead. That is design information and falls outside the limit; dimensions, mass and published performance claims fall inside it.
We write here that we read these fields and left them out. Skipping them silently would mean reporting the source incompletely. The link to the source is on the page already, and readers can consult the document themselves.
Corrections
If you believe a value is wrong, or you hold a better source, send it. Every sourced correction goes on record. Contact details are on the about page.