Philip Tetlock spent 20 years tracking 28,000 predictions from 284 expert forecasters. He found that the average expert was about as accurate as a dart-throwing chimp.
But a small group — roughly 2% of his sample — was different. They were dramatically better than chance, and they kept getting better year after year. He called them superforecasters.
The good news? They were not born with it. Tetlock's research shows the skills are learnable. Here's how to actually get better at predicting the future.
Step one: Forecast in numbers, not words
Most forecasts use weasel words: "likely," "possible," "significant chance." Research shows two people reading "likely" can mean anywhere from 30% to 90%. The language is a fog.
Force yourself to use a number. Not "I think US GDP will probably beat 2%." Say "I think there's a 65% chance US GDP beats 2%."
This single change does three things. It forces clarity. It creates a track record you can score. It makes you confront your own confidence.
Step two: Start with the base rate
Before you look at the specific question, look at the general one.
"Will the next Fed meeting cut rates?" Don't start with the latest news. Start with how often the Fed has cut rates in any given meeting over the past 10 years. That's your base rate.
If the base rate is 15%, your forecast should start at 15% and adjust from there. Only move significantly if you have a specific reason — a slowing economy, weak jobs data, a credit event.
Most people skip the base rate entirely. That's why they overreact to news.
Step three: Break the question into smaller pieces
"Will US GDP beat 2% in 2026?" is too big. Break it into the moving parts.
What's the contribution from consumer spending? From business investment? From net exports? From government spending?
Forecast each piece separately. Then sum them up. You'll often find that the headline forecast was driven by one or two underlying assumptions you didn't notice.
This is how the best forecasters get their edge. They don't think harder about the big question. They reframe it into smaller ones.
Step four: Update slowly, but actually update
Bayesian thinking, without the math, is this: when new information arrives, move your forecast in the direction the information points — but not too far.
The hardest part is the second half. Most people either ignore new information or overreact to it. Superforecasters do neither.
A useful rule of thumb: if the news doesn't change the base rate, it shouldn't change your forecast much. A single tweet from a politician is rarely worth 10 percentage points. A 50-basis-point Fed move usually is.
Step five: Keep score
Tetlock's superforecasters did one thing almost nobody else does. They wrote down every prediction and graded themselves afterwards.
Keep a list. Every time you make a prediction with a probability attached, log it. Every time it resolves, score it.
After 50 predictions, you'll know exactly where your blind spots are. After 200, you'll start getting noticeably better. After 500, you'll be in the top 10% of forecasters in the world.
Step six: Surround yourself with disagreement
The fastest way to get worse at forecasting is to only talk to people who already agree with you.
Superforecasters actively seek out the strongest version of the opposite view. They steelman it. They ask "what would have to be true for the other side to be right?" — and then they look for that evidence.
If you can't find anyone who disagrees with you, the problem is not the question. It's your information diet.
Step seven: Use markets as a sanity check
Prediction markets are not magic. They are just very fast aggregators of everyone else's forecasts, weighted by how much money those people are willing to risk.
When the market price disagrees with your forecast, that's a signal. Either the market is wrong and you've found an edge — or your forecast is wrong and you need to update.
The honest answer, most of the time, is the second one. Use the market to find your blind spots.
What Juno lets you practice
Juno is a prediction market designed to let you do exactly this. Forecast a number. Watch the market price update. Score yourself when it resolves.
It's the only way to actually get better at forecasting — by treating it as a skill you practice, not an opinion you share.
The 2% that beat the chimp
Tetlock's superforecasters were not smarter than the average expert. They were not better educated. They didn't have special information.
They just did the seven things above, consistently, for years.
Start with one of them this week. The track record will take care of itself.