Every dated battle and siege Wikidata knows, with its body count where history recorded one: 16,961 engagements spanning 1500 BCE to the present, BCE years shipping as negatives. The roster walks all 34 subclasses of battle and siege (naval battles, air battles, pitched battles, skirmishes) and accepts both date forms, because requiring only point-in-time dates silently dropped the deadliest battles in history, the multi-month Stalingrads and Sommes that carry start dates instead; a regression test now guards Stalingrad's presence forever. Wikidata itself records deaths on barely 131 battles, so the body counts are a hybrid: the graph's figure where present, else mined from each Wikipedia article's Casualties-and-losses infobox cells, lifting coverage to 4,782 battles, each tagged with its Deaths_Source and both sides' raw casualty text preserved for judgment. Rows carry the war, era, battlefield coordinates (59%), participants, and siege durations in days. The deadliest in the data: the Battles of Rzhev, 1942, at 2.86 million. A missing body count is history, not a zero.
16,961 rows in one CSV file, with 15 columns in total: Battle, War, Year, Date, Era, Country, Latitude, Longitude, Deaths, Deaths_Source, Casualties_Side_A, Casualties_Side_B, Participants, Duration_Days, Battle_URL.
It is built by programmatically scraping wikidata.org, last pulled on 2026-09-12. Datasets are versioned; older versions stay downloadable.
Yes. Every CodeSights dataset is completely free as a CSV download; a free account is all it takes. Anyone can preview the data without signing in.
Yes. The exact Python scraper that built it is viewable on the dataset page by any signed-in member, so every number is reproducible.
Automated scraping leaves room for error and the underlying sources change over time, so no version is guaranteed accurate or complete. If a number matters, verify it against the original source.