Where self-assessment breaks, and in which direction

Ask a room of drivers to rate themselves against the average driver in the room. More hands go up for "above average" than can possibly be right. Ask the same room about their ability to juggle and the effect vanishes.

That contrast is the actual lesson, and it is more useful than the cliché. Self-assessment does not fail uniformly. It fails in a specific place, in a specific direction, and knowing where lets you predict which of your own judgements to distrust.

Three separate things wearing one name

"Overconfidence" gets used for three effects that behave differently and need different corrections.

Overplacement is thinking you rank higher than you do. This is the driving one. It shows up on easy, common tasks and reverses on hard, rare ones — most people correctly suspect they are below average at juggling, and that self-doubt is not humility, it is feedback.

Overprecision is being too sure of a number. Asked for a range you are 90% confident contains the right answer, most people give a range that contains it far less often. This is the one that costs money, because every financial plan is a range wearing the costume of a point estimate.

Overestimation is thinking you will do better than you will — finish sooner, spend less, earn more. It survives being told about it, which is the irritating part.

Predict

Which of the three is most directly responsible for a renovation that was quoted at six months and takes fourteen?

Choose one answer

Overplacement — the people involved thought they were better than other buildersOverprecision — the six months was a point estimate with no range attachedOverestimation — the plan was built from a best case that nothing intervened inNone of them — the builder was being dishonest to win the work

Lock in answer

Why it is not vanity

The tempting explanation is ego. Ego is involved, but it is not necessary, and the effect appears in places where there is no audience to impress.

The structural reason is that you have access to your own reasoning and to nobody else's. You can see the care you took, the things you considered, the reasons behind your decision. Of everyone else, you can see only the outcomes. So your process is compared against their results, and your process always looks more thorough — because for them there is no process visible at all.

The same mechanism runs in the other direction in the one situation where you can see somebody's process: you tend to be far more confident about a stranger's ability than a colleague's, and the difference is not usually that the stranger is better.

Not the same thing

Confidence

How sure you feel about a judgement.

Measured in

A stated probability, or a range you would bet on

For example

“I am about 80% sure this project ships this quarter.”

Chase it alone and

On its own it is neither good nor bad — it only becomes a problem when nothing checks it against outcomes.

Calibration

Whether things you are 80% sure of actually happen about 80% of the time.

Measured in

Hit rate against stated confidence, across many judgements

For example

Twenty predictions made at 80%; sixteen of them come true. That is calibrated.

Chase it alone and

It cannot be assessed from any single judgement, so people substitute how confident they felt, which is the thing being tested.

Which is which? Put each one on a side.

• Feeling certain about a forecast

  ConfidenceCalibration

• Keeping a written record of predictions and checking them a year later

  ConfidenceCalibration

• Being right about something difficult

  ConfidenceCalibration

Where it shows up in money

Three places, roughly in order of cost.

Trading frequency. Believing you can tell which of two things will do better is a belief about your rank against everyone else pricing them. Acting on that belief costs money on every transaction, whether or not the belief is true, which makes it an unusually expensive thing to be wrong about.

Concentration. A portfolio holding one thing is a statement that you know something specific. Sometimes that is true, and concentration is genuinely how most large fortunes were built. It is also how most of them were lost, and the difference is not visible from inside.

Plans. Every projection in this course — including all the models — is a point estimate that will not happen. Building around one number and no range is overprecision expressed as a spreadsheet.

Check

Which correction actually improves calibration, rather than only feeling like humility?

Choose one answer

Reminding yourself that you might be wrong before decidingWidening every range by a fixed amountWriting down predictions with dates and confidence levels, then scoring them laterAsking other people whether they agree

Check answer

The version worth keeping

None of this argues for having no opinions. Someone who is uncertain about everything cannot act, and a bias towards believing you can do the thing is probably why anyone ever starts anything difficult.

The distinction that survives is between confidence in a decision and confidence in a forecast. Being sure you want to take a course of action is a statement about your preferences, and you are the authority on those. Being sure about what will happen is a statement about the world, and the world does not care how carefully you thought.

Worth remembering

• Self-assessment is calibrated by feedback, so it drifts upward wherever feedback is slow, noisy or ambiguous.
• Overplacement, overprecision and overestimation are three different effects and need different corrections.
• Estimates run late far more often than early, and an asymmetric error is the signature of a bias rather than noise.
• You can see your own reasoning and only other people's outcomes, which makes your process look more thorough than theirs.
• Calibration is a property of a track record, and it cannot be assessed from any single judgement.