How to Make Better Predictions

Good forecasting is not about being right more often. It is about being
calibrated: when you say you are 70% sure, you should be right about 70%
of the time.

Start with a base rate

Before you reach for the specifics of a question, ask: how often does this
kind of thing happen in general? A startup's odds of surviving five years,
a project shipping on time, a habit sticking past a month — each has a base
rate. Anchor on it, then adjust.

Break confidence into numbers

Probably is useless feedback. Force yourself to a number: 60%, 85%, 30%.
Numbers can be scored; vague words cannot. Over time, scored numbers are the
only way to learn whether your gut is trustworthy.

Update in small steps

New evidence rarely justifies jumping from 50% to 95%. Move in increments.
The forecasters who improve fastest are the ones who nudge their estimates
as evidence trickles in, rather than swinging wildly.

Review your misses

The point of writing a prediction down is to look back. Once outcomes are
known, revisit your confident wrong calls. Patterns — overconfidence in a
domain, anchoring on recent news — show up only when you keep score.