Where Was the Battle? Testing Proposed Sites of Lost Early Battles in Britain against their Texts

Many of the best-known battles of Roman and early medieval Britain have no known site. I assembled 88 places that have been proposed for 18 such battles, from Boudica's final battle (AD 60 or 61) to Assandun (1016), with a well-located control, Heavenfield (634). Each proposal keeps its source, and each point comes from a cited gazetteer, grid reference or coordinate; none is guessed. For every battle I list only the clues the early texts give about the place (a hill, a siege, ships, a river, a lake, a named place nearby), turn each into a check with an explicit threshold, measure the ground at every site from open elevation, river, road and coastline data, and score each site by the share of clues it meets. The texts give only 31 testable clues across the 18 battles, and two battles give none. Of 85 scored sites, 20 meet every clue for their battle and 4 meet none; in 9 of 16 scored battles two or more sites tie for the best fit. The theories are far apart: for the 15 lost battles with more than one proposal, the median distance between the two most distant sites is 89 km, and Camlann's proposals span 679 km. The tightest cluster is the Battle of Whalley (798) in Lancashire, whose five proposed sites lie within 3.5 km. Wikidata gives a single coordinate to 12 of the 17 lost battles, two of them more than 20 km from any proposed site. The early texts rarely choose between candidates; they rule a few out. Interactive version: https://tristan-hotham.github.io/where-was-the-battle/. Data and code: https://github.com/Tristan-Hotham/where-was-the-battle. Blog post: https://blog.tristanhotham.com/2026/10/27/where-was-the-battle/.