One row per shipwreck Wikidata can tell a story about: 3,501 lost vessels across 130 countries and four millennia, from the Bronze Age Dokos wreck (sunk around 2150 BC, negative years mean BC) to losses reported this year. The roster walks every wreck route in the graph: the shipwreck class tree, sinking, scuttling and capsizing events, ambiguous events like torpedo attacks that only count when the ship also shows an end (a torpedoed ship can survive; the Wilhelm Gustloff did not), and standalone disaster items linked back to their hull. Then each English Wikipedia article's ship infobox fills what the graph left NA: the Fate line as written on 2,435 rows, casualties parsed from it conservatively, launch and sinking dates, tonnage. 3,144 wrecks carry a sinking year and 2,655 know the exact day, so the wars tower over everything (214 losses in 1914-1918, 1,559 in 1939-1945) and seasonality is computable. Alongside the dates: launch year with years afloat (2,683 lifespans), cause of loss, death tolls that follow the link to the disaster event when the ship item is silent (the Titanic's 1,501 dead live on the event), survivors, gross tonnage and displacement, dimensions normalized to metres however they were entered, wreck coordinates for 2,212 hulls with body of water and depth, and the year the wreck was found (the Titanic in 1985). Honesty guards: no length exceeds the 458 m the largest vessel ever built actually measured (larger claims are graph typos and go NA), a 1969 grounding cannot become the cause of a 1975 sinking, and a multi-vessel disaster's toll belongs to no single hull. A row earns its place by having a story (an article, a date, a death toll or a tonnage): the graph's 29,654 coordinate-only seabed survey points and 308 non-vessels that ships crashed into (the Francis Scott Key Bridge, the Suez Canal) are dropped and disclosed. Built with a 13-test acceptance suite anchored on the Titanic's exact night and tonnage, the Vasa's 1628 capsizing, the Wilhelm Gustloff's torpedo attack and the visibility of both world wars.
3,501 rows in one CSV file, with 27 columns in total: Ship, Ship_ID, Ship_Type, Country, Home_Port, Operator, Builder, Conflict, Launched_Year, Sunk_Date, Sunk_Year, Years_Afloat, Cause, Fate, Deaths, Survivors, Gross_Tonnage, Displacement_Tonnes, Length_m, Beam_m, Draft_m, Latitude, Longitude, Body_of_Water, Wreck_Depth_m, Wreck_Found_Year, Ship_URL.
It is built by programmatically scraping wikidata.org, last pulled on 2026-09-15. 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.