In 1984, a psychologist named Philip Tetlock started a 20-year study on expert prediction. He tracked thousands of forecasts from hundreds of experts. His finding: most experts predicted the future about as well as a dart-throwing chimp.
But not all of them. A small group consistently got things right. He called them superforecasters. And what they did differently is teachable.
What Separates Good Predictors from Bad Ones
Bad predictors have strong opinions they never update. They're confident. They explain away their failures. They never track their results because that would force them to confront how wrong they are.
Good predictors think in probabilities. They update constantly. They're calibrated — meaning when they say "70% chance," they're right about 70% of the time. They're not afraid to say "I don't know."
The goal isn't to be right. It's to be calibrated. There's a difference — and it matters a lot.
The Five Habits of Superforecasters
- →Break big questions into smaller, answerable sub-questions
- →Start with base rates before incorporating specific information
- →Actively seek out evidence that contradicts your view
- →Update your probability estimates as new information arrives
- →Track every prediction and review your accuracy over time
Base Rates: The Most Underused Tool
"Will this startup succeed?" Most people answer based on the specific startup. Superforecasters start by asking: "What percentage of startups at this stage succeed?" That's the base rate. Then they adjust based on specifics.
Base rates are boring. They're also nearly always right, and nearly always ignored.
How to Practice
Forecasting is a skill. Like any skill, you improve by doing it repeatedly with feedback.
Prediction markets are the best practice environment ever invented. You make predictions, you receive immediate feedback through price changes, and you get paid when you're right. The feedback loop is tight and real.
The Honest Truth About Predicting the Future
Nobody consistently predicts the future with certainty. That's not the goal. The goal is to be slightly better than the market on average, over many trades.
"Slightly better than average, over many trades" sounds boring. Compounded over time, it's how people make real money in prediction markets.
The chimp doesn't have an edge. You're trying to get one.