war atlas
5 papers · read online or download the PDF
Short studies of warfare, supply, and settlement. Each begins with a question about how the past worked — whether bombing a city made it harder to capture, how much of an army’s supply train was food for its horses, whether crossing more open ground cost an attacker more men.
Paper 01
2026-09-09
result publishedfigures pending source verification
A descriptive study with a decision rule stated before counting
Four large frontal assaults conducted in the American Civil War between July 1862 and July 1863 are compared on two quantities: the approximate distance of open ground the attacking formation had to cross under fire, and the share of the attacking force lost. Across the four cases, loss share rises with distance crossed, from about one…
Paper 02⚠ number clash
2026-09-09
published refusalfigures pending source verification
A registered study that could not be run, published as a documented absence rather than withdrawn
A study registered to compare ordnance delivered onto four cities against how hard each proved to take is refused, because neither variable exists in the corpus searched. Zero of four ordnance figures are present; the outcome variable has no operational definition, no column and no scale; and two of the four named cases have no database…
Paper 02⚠ number clash
2026-09-09
result publishedfigures pending source verification
A diagnostic of the acknowledgement field across eighteen proxy conflicts, and of the count already published over it
A single column recording whether an outside sponsor admitted sponsoring a combatant holds eight distinct free-text formulations across eighteen conflicts. A control column in the same sheet, by the same compiler, partitions the same rows into five clean terms, so the spread is a property of the concept rather than of the compiler. The…
Paper 03
2026-09-09
published refusalfigures pending source verification
The ratio cannot be computed. What can be computed instead refutes the source that asked for it.
The briefed ratio — fodder as a share of total supply tonnage for a named pre-mechanised force — cannot be computed from any source on this machine, because neither the denominator nor the force composition exists. The refusal is published. What the rows do support is a class 1 corpus diagnostic over the complete 18-row logistics_rates…
Queue item 5
2026-09-09
published refusalfigures pending source verification
A registered claim, refused rather than tested — and the corpus diagnostics that made the refusal necessary.
A registered claim — that the physical logic of a city's site predicts how long the city persists — is refused rather than tested. The corpus that would test it is the corpus the claim was written from, and it carries no persistence variable, no admission rule, and one source across all 59 rows. What is published instead is a class 1…
Measure class
A measure class says what a number is a statement about: class 1 describes one named table or file and nothing outside it, class 2 compares a handful of named cases and claims nothing beyond them, and class 3 asserts something about a whole population — which is why a class 3 paper cannot be published without a written rule saying who was eligible to be counted before anybody counted.
Status
Either the paper reached a result it is prepared to defend, or it is a refusal — a study that was registered, attempted, and then published as a documented absence because the variable it needed turned out not to exist. The listing marks which, before you open anything.
Figures pending
Figures pending source verification means the method is finished and the numbers in the table have not yet been traced back to a named published source. The papers ship in that state on purpose: the method is the part being offered, and the figures are the part a reader can improve.
how this page is built
This listing is read off the directory the papers live in, and every class, date and status above is parsed from the paper file itself. Nothing is transcribed, so nothing can go stale against the paper it describes. 2 papers currently declare the same paper number; the listing shows the clash rather than renumbering them, because a number changed here would disagree with the PDF a reader downloads.
