On the source. Borders, statuses and dates come from CShapes 2.0 (Schvitz, Girardin, Rüegger, Weidmann, Cederman and Gleditsch, “Mapping the International System, 1886–2017”, Journal of Conflict Resolution 66(1), 2022). It is published under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 licence, and the simplified file this page loads carries the same terms — which is why the map you can save from this page prints that attribution into the image. Nothing on this map is drawn by hand.
On what counts as a change. The dataset records changes in law, not on the ground. A treaty signed is the day a border moves; a front line is not. It also leaves out territorial changes under 10,000 km², unrecognised secessions, and conquests reversed by the end of the same war. That is why Poland does not vanish in September 1939 and why Manchukuo never appears. Those are wartime occupations, and this page draws them as a hatched overlay in the occupier’s colour, with its own dated sources — never by editing the borders underneath.
On Crimea, the one correction. The source dataset records the accession treaty of 18 March 2014 as a border change and draws Crimea as Russian from that date. This map does not. The dataset’s own coding rules exclude changes that lacked international recognition, and nine days later the UN General Assembly declared the referendum behind it invalid. So the borders here stay as every other state recognises them, and the Russian occupation is drawn the way this page draws every other one — as hatching, from March 2014, with its sources. It is the only place this map departs from its source, and the build script names it.
On the end of the record. The dataset stops on 31 December 2019. That is not an omission — no border recognised in international law has changed since. The wars of the 2020s in Ukraine, Gaza, Sudan and Nagorno-Karabakh have moved front lines, populations and control, none of which this map is drawing yet.
On the writing. The change log beside the map is generated from the dataset and nobody writes it. The chapters are the opposite: every event, every occupation overlay and every hovered sentence was written by hand, and each one carries at least one source that is not an encyclopaedia — treaty texts, the Nuremberg judgment, court records, and the archives of the United Nations and the US State Department. A script in this repository checks both that every link resolves and that each cited page is actually about what it is cited for, since a page can answer with the right title and never mention the subject. Where an entry’s only source is the dataset, that source vouches for its date and its change, not for every word around them. Many of the hovered sentences state a count, and a test recomputes each one from the map. Entries marked contested are live or disputed, and are held to a deliberately narrow vocabulary; read their sources before their summaries.
On the quiz. No question is written by hand. Each one is generated from a period this dataset records or an event on the page, so the quiz cannot claim anything the map does not, and an event question carries that event’s sources into the answer and into the sheet you export. While a question is live the map stops colouring by ruler, because the colours are the answer to half of them.
On the years before 1886. The five stones to the left of the timeline are a different kind of map and are labelled as one. They come from the historical-basemaps project, which describes its own work as in progress and historical boundaries as more disputed than modern ones, and they carry no dates at all — only the project’s own grading of how firm each boundary is, which you can see on any shape by hovering it. They are published under the GPL, not the licence above. No quiz question is ever drawn from them.
On the colours. Territories are coloured by the power that held them, with independent states sharing one neutral tone. Shade carries the kind of control — a protectorate is lighter than a colony, a mandate lighter still, and occupation is hatched. Names are the dataset’s own and lean modern: the British colony that became Malaysia in 1963 is labelled “Malaysia” from 1946.