Rounds
Ten questions in about a minute. One ranked round a day, the same for everyone, with a leaderboard; then as many quick rounds as you like. Challenge a friend to the same ten.
Who's bluffing?
Ten questions. Stake how sure you are. The scoring rule knows when you're bluffing.
About a minute. Anonymous — no account, no tracking. Your answers go into a public research dataset.
A new ranked round every day at 00:00 UTC
How it works
Ten questions, about a minute
Pick one
Each question compares two things: which river is longer, which mountain is higher, which plane flew earlier.
Say how sure you are
From 50% (a coin flip) to 100% (stake it all). Honest confidence scores best: a sure miss costs 300 points, a coin flip costs nothing.
See how you did
Both answers and their sources appear right away. At the end: your score, your type, and how sure you were against how often you were right.
Three ways to play
On your own, with your team, or in class
Slack
Install once. One question a day in a channel: members pick an answer and say how sure they are, and the post later reveals the answer and the boldest bluff.
Add to SlackClassroom
Run the overconfidence demo in class. Students play on their phones while your screen shows the class’s calibration curve.
For teachersOpen research
A game, and a study
Does daily feedback make people better at knowing how much they know? The hypotheses are fixed before launch, the anonymous answers will be released as open data, and all the code is public.
FAQ
Questions
What does “how sure” mean?
Your confidence is your chance of being right. Say 70% on ten questions and you should get about seven of them right. Most people are surer than they are right; the game shows you by how much.
How are points worked out?
Each answer scores 100 − 400 × (confidence − outcome)², where the outcome is 1 if you were right and 0 if not. So 50% always scores 0, 100% scores +100 when right and −300 when wrong, and 80% scores +84 or −156. The best strategy is to say exactly how sure you are; the research page explains why.
What do the types mean?
Your type compares how sure you were with how often you were right: Bluffer (15 or more points surer than right), Hot-headed (5 to 15 points surer), Calibrated (within 5), Modest (5 to 15 points less sure) and Hedger (15 or more points less sure).
What is the ranked round?
The same ten questions for everyone, new every day at 00:00 UTC, with a leaderboard. After it, quick rounds pick questions you haven’t seen yet. Your streak counts the days you finish at least one round; missing a single day doesn’t break it.
Do I need an account?
No. There are no accounts and no tracking. Your browser keeps a random id so your streak survives a reload; we never store your IP address or an email. If you add a nickname to a challenge link, it is shown to the people who open that link. Details are in the privacy policy.
Where do the questions come from?
From Wikidata, with a source link for both answers. Two things are compared only when one is at least 1.3 times the other (for years, at least ten years apart). Ranked rounds use only facts with a reference or an independent check, and you can report a question that looks wrong right under its answer.
Is there a longer version?
Yes. The full assessment takes 5 minutes: twelve two-choice questions with a confidence rating and six 90% ranges, then a chart of how sure you were against how often you were right.