They are a model, not a room of people we phoned. Every count in this game is our careful estimate of how 1,000 American respondents would answer that question cold — built from published research on what people name first in a category, then checked by hand, answer by answer.
That is why the game never says “we asked 1,000 people”. It asks “how would 1,000 people answer?” — because that is the honest question, and the numbers are our best answer to it. Which 1,000 people, and why it matters, is the next section.
A count only means something against a specific crowd. Ask the world and you get mush; ask a defined panel and rarity becomes real. We picked one and stuck to it, so “rare” has a consistent meaning from day to day.
Answers from anywhere else are completely valid — they simply tend to live in the tail, which is exactly where the points are. Naming something 1,000 Americans would not think of is the whole game.
After you finish a day, we ask for your first answer to a question you have not played yet. That answer is cold, so it is real frequency data — and once a question has enough of them, the model is thrown away and the real counts take over. The reveal then carries a badge:
There is no opposite badge. A day without one is simply a day still running on the model, and we would rather say nothing than imply something.
We never use your gameplay guesses as data. By the time you are three disks up you are hunting for rare answers, not saying what comes to mind — those guesses measure your tactics, not the crowd.
Then tell us. Every answer on the results screen has a ⚑ beside it. Flags are reviewed weekly, and the counts that players argue with most are the first ones we re-check.