Overconfidence is the default, not the exception
Across decades of studies, when people are asked to give a range they're "90% sure" contains some true value — the length of a river, the year of an invention — the true number falls inside their range far less than 90% of the time. Researchers have found this pattern repeatedly, across topics and across levels of expertise: the ranges people offer tend to be too narrow, reflecting more certainty than the evidence supports. (These are well-known results in the judgment-and-decision-making literature. The exact percentages vary by study and question set, so treat any specific figure you see quoted online with care and check the original source.)
Overconfidence isn't a sign of stupidity or arrogance. It shows up in experts and beginners alike, and the conditions you'd expect to fix it — knowledge, high stakes — don't reliably do so on their own. What does help is feedback that is fast, clear, and repeated. That last point is the crux, and it's where keeping a record comes in.
Why you can't feel it
Two ordinary features of the mind keep overconfidence hidden. First, confidence is a feeling, and feelings aren't calibrated to accuracy — the sensation of being sure is the same whether the real odds are 95% or 55%. Second, hindsight: once you know how something turned out, it becomes hard to reconstruct how uncertain you really were beforehand. "I knew it all along" erases exactly the data you'd need to grade yourself. Without a record written before the outcome, there's almost nothing to check your confidence against.
A quick self-check — and its limits
There's a classic exercise you can do in a couple of minutes. Answer ten trivia questions, but instead of a single guess, give a range you're 90% sure contains the answer (for example, "the Nile is between 5,000 and 8,000 km long"). If you were perfectly calibrated, about nine of your ten ranges would contain the true value. Most people land closer to five or six — a rough, immediate hint that their "90%" behaves more like "60%".
It's a good demonstration, but be honest about what it is: a one-off snapshot, on trivia, on a tiny sample, that you can game the moment you learn the trick (just widen every range). It can nudge you to notice the tendency. It can't tell you whether you're overconfident about the things that actually matter to you — your work, your plans, your read on people — or by how much, or whether you're getting better. For that, you need a different kind of test.
The real test: track your own predictions
The reliable way to answer "am I overconfident?" is to measure your own forecasts about your own life, over time. It's simple in principle: when you make a call, write it as a specific, checkable claim, attach a confidence from 1% to 99%, and set a date to check back. When the date arrives, mark it right or wrong. Do this a couple of dozen times and a pattern appears that no quiz can give you.
Group your predictions by the confidence you assigned. If the things you called "90% likely" happened about 90% of the time, you're well-calibrated there. If they happened 60% of the time, you're overconfident — and now you know it as a number, not a worry. The same record catches the opposite too: plenty of people are systematically under-confident, hedging on calls they actually get right. One summary figure — the Brier score — rolls all of this into a single number you can watch move. For how that score is calculated, see What is calibration? The Brier score explained.
What to do once you know
Finding out you're overconfident isn't a verdict on your intelligence — it's one of the more useful things you can learn about your own judgment, precisely because it's fixable. The fix is boring and effective: on the kinds of calls where your record shows you miss, drop your confidence a notch and widen your ranges. People who get regular, honest feedback on their forecasts tend to become better calibrated over time — not because they get smarter, but because they stop mistaking the feeling of certainty for the fact of it. The one requirement is a record you keep honestly — ideally one only you can see, so there's no temptation to look good rather than be accurate. (On why privacy matters for honest self-assessment, see Why track predictions privately?)
One tool built for this
Hunch is a free prediction journal that runs entirely in your browser — no account, no sign-up, and your predictions never leave your device. You log a prediction with a confidence and a resolve-by date, mark it correct or incorrect later, and it shows your calibration and Brier score, flagging whether you lean over- or under-confident. It won't hand you a verdict in two minutes — nothing honest can — but it's built to turn "I think I'm a good judge" into something you can actually check. If you'd like to compare it with other tools first, see Hunch vs PredictionBook vs Fatebook, or browse more reading in Resources for calibration & forecasting.