Based on Paul Nation’s BNC/COCA word family lists, which rank the most frequent 25,000 English word families using data from the British National Corpus and the Corpus of Contemporary American English. This site uses the first ten 1,000-word bands, via the machine-readable version published at github.com/lpmi-13/machine_readable_wordlists.
Adapted from the Wordset dictionary, derived from Princeton University’s WordNet. Licensed under CC BY-SA 4.0.
Words are sampled evenly across ten frequency bands. Your accuracy within each band is extrapolated to estimate how many of that band’s 1,000 word families you know, and the bands are summed for a total estimate. This mirrors the approach used by academic vocabulary size tests such as Nation & Beglar’s Vocabulary Size Test.