A little science behind the guesses.
In Guess the Rarest Word, you pick the word that appears less often in our English frequency dataset. Every pair uses the same data, so the answer follows a consistent rule.
How to play
Choose one of the two cards, or press 1 or 2. A correct answer earns 100 points. Consecutive correct picks add a bonus of 20 points per extra win, up to 100 bonus points per pick. Each wrong answer costs a heart. Your run ends when all three hearts are gone.
Classic gives you 12 seconds for each pair after the first warm-up. An expired timer costs a heart. Chill lets you take your time. The timer pauses when you leave the browser tab, open instructions, or use the word comparator.
Where the numbers come from
We use the English large list from wordfreq 3.1.1 by Robyn Speer. It combines books, subtitles, news, Wikipedia, web text, Twitter and Reddit. This is a snapshot of usage through roughly 2021.
Each word has a frequency estimate on the Zipf scale. A score of 3 means about one occurrence per million words. A score of 4 means about ten per million. We display the estimates per million words to make the comparison easier to read.
What counts as a word?
The comparator checks individual written words. Uppercase and lowercase are treated alike. Plurals, tenses and different spellings have their own entries. Meanings that share the same spelling are counted together. Accents are preserved.
Some words have the same rounded frequency. We call those a tie. If a word is missing from this dataset, the comparator says it needs more data. A missing entry has no reliable frequency estimate here.
The game uses 190 original word pairs, selected and checked against this fixed dataset. Frequency measures how often a word appears in the source material. Familiarity and difficulty can vary from person to person.
Credits
Word-frequency data © 2022 Robyn Speer, adapted from wordfreq 3.1.1. Our converted data and frequency scores are available under CC BY-SA 4.0. Wordfreq software is licensed under Apache-2.0. The data includes freely available SUBTLEX-US by Marc Brysbaert and Boris New, SUBTLEX-UK by Walter J. B. van Heuven, Pawel Mandera, Emmanuel Keuleers and Marc Brysbaert, and OpenSubtitles.
Full credits and conversion notes · Original data notices · Data license
This is an independently made browser game. Its word pairs, artwork and frequency scoring are created for this site. It is not affiliated with Roblox or the Roblox experience of the same name.