← Back
🪙

Confidence Bet

Solve the problem and bet
on how sure you are.

🏆 Best reserve0
🔥 Daily streak0 days
NumCalcX

How Confidence Calibration Works

Calibration is the gap between how sure you say you are and how often you're actually right. A perfectly calibrated person who says "80% confident" on a batch of answers will turn out to be correct on close to 80% of them — not 95%, not 60%. Confidence itself isn't inherently good or bad; what matters is whether it tracks reality. Psychologists have documented overconfidence as one of the most consistent findings in judgment research: across many domains, people systematically state higher confidence than their actual accuracy supports, especially on harder questions where they have the least reliable information to go on.

This game measures that gap directly. Every bet you place is a confidence statement — betting 90% of your reserve says "I'm nearly certain," betting 10% says "I'm mostly guessing." After 10 rounds, the calibration chart groups your bets into confidence buckets and plots how often you were actually correct at each one; the diagonal line is what perfect calibration looks like, and the diagnosis compares your average stated confidence against your average real accuracy to tell you whether you tend to overbet, underbet, or track yourself accurately.

Frequently Asked Questions

What does it mean to be "well calibrated" in this game?

It means your average bet size roughly matches your average accuracy — if you're right about 60% of the time, a well-calibrated player bets around 60% on average. It's not about being right often, it's about your confidence being an honest signal of how likely you actually are to be right.

Why does betting low on an answer I'm actually sure of cost me points?

Because the game only rewards confidence that turns out to be justified. If you're consistently right but bet cautiously anyway, you're underconfident — you leave points on the table by not trusting information you actually have. The diagnosis treats that as a real, measurable calibration problem too, not just a safe strategy.

Is overconfidence a real, documented bias, or just something this game made up?

It's real and well studied in psychology and decision research, particularly on questions where people have incomplete information — stated confidence tends to run higher than actual accuracy. This game recreates a simplified version of that pattern using math and physics problems instead of general-knowledge quiz questions.