Overconfidence is quietest when it is most expensive

There is a settled way of talking about overconfidence. It is worth asking how much of it survives contact with the detail.

The general finding

Across many domains, people rate their own abilities above the median more often than is arithmetically possible. In investing this appears as confidence about timing, about selecting funds, and about the ability to stay calm in a fall. The third is the most consequential, because it determines how much risk people take on.

Nobody discovers whether it was true until conditions test it.

Confidence outpaces accuracy

Additional information and additional experience both reliably raise confidence; whether they raise accuracy depends on the quality of feedback. Investing has slow, noisy feedback, so experience accumulates without much correction.

Put simply, ten years of investing therefore produces a decade of confidence and relatively little verified skill. That is not a criticism of investors; it is a property of the feedback environment.

Memory does the rest

Predictions that turned out well are recalled clearly and those that did not are revised or forgotten. The result is a personal track record that is systematically better than the real one. This happens without dishonesty and is very difficult to notice from the inside.

Anyone who believes they remember their forecasts accurately has not written any down.

The correction is a notebook

Record predictions with dates: what you expect, why, and how confident you are. Reviewing them a year later is the only way to see your actual hit rate rather than the remembered one. Most people who do this stop making predictions, which is the intended outcome.

It costs nothing and is the single most effective humility mechanism available.

How overconfidence spends money

It shows up as larger positions, more concentration, more trading and less diversification. It also shows up as choosing an allocation based on an assumed tolerance for loss that has never been tested.

In practice, studies of individual investors have generally associated higher trading activity with lower net returns, which is consistent with confidence exceeding skill. The costs are ordinary rather than dramatic, which is why they accumulate unnoticed.

Useful confidence

Confidence in a process, in continuing to contribute and in holding through declines is different from confidence in judgement about markets. The first is a commitment you control and the second is a claim about the world. Directing self-belief at behaviour rather than at prediction is both more accurate and more useful.

For most people, for any decision material to your circumstances, regulated advice locally is a better check than self-assessment.

Calibration is the measure that matters

Being right often is not the same as being well calibrated, which means your stated confidence matches how often you turn out to be correct. Someone who says they are almost certain and is right four times in five is badly calibrated despite a good hit rate, because the confidence was wrong even where the call was not. Calibration is trainable in fields with fast, clear feedback — weather forecasting is the standard example — and largely untrainable in fields without it, which is the position investing is in.

The practical consequence is to attach a probability rather than a verdict to anything you expect, since a claim of near-certainty about a market is almost always the confidence talking.

The takeaway

Write your predictions down with dates. Your remembered record is not your record.

Pick the one that costs you least, and let the rest wait.