Get early access

Methods4 min read

What is a superforecaster? Who they are and how to become one

Superforecasters are the top forecasters from a US intelligence tournament. Who they are, what makes them accurate, and how to become one.

By Kristian FagerlieFounder & CEO, Sikt Intelligence

A superforecaster is someone whose forecasts of world events are consistently and measurably more accurate than other people's. The term comes from a four-year US intelligence tournament, where the best amateur forecasters, roughly the top 2%, beat professional intelligence analysts. Their secret was not genius or inside information. It was a set of habits anyone can learn, and that are now being built into AI.

Where superforecasters come from

In his 2005 book Expert Political Judgment, the psychologist Philip Tetlock showed that the long-range forecasts of many pundits and experts were often little better than chance. The obvious follow-up question: can anyone forecast well, and how?

In 2011, IARPA, the research arm of the US intelligence community, launched a forecasting tournament to find out. Over four years, 500 questions and more than a million forecasts, the Good Judgment Project, led by Tetlock and Barbara Mellers, won. Its best forecasters beat the other research teams by 35–72%, and according to Good Judgment they were more than 30% more accurate than intelligence analysts with access to classified information (Good Judgment). That top group became known as superforecasters.

What makes a superforecaster

Tetlock and Dan Gardner described the research in Superforecasting: The Art and Science of Prediction (2015). The superforecasters were not a type of person so much as a way of working:

  • They think in probabilities. Not "it will happen", but "about 30%", and they distinguish 60% from 70%.
  • They start from the outside view. Before the details of this case, they ask how often similar things happen: the base rate.
  • They break questions into parts. A hard question becomes several easier ones.
  • They update often, in small steps. New evidence moves the number, but rarely all at once.
  • They are actively open-minded. They look for reasons they might be wrong, and change their minds.
  • They keep score. They know their track record, and learn from their misses.

The ten commandments, in short

The book ends with "Ten Commandments for Aspiring Superforecasters" (summary). A few of the most useful:

  1. Triage. Spend effort on questions where it pays off, not on the trivially easy or the truly unknowable.
  2. Break seemingly intractable problems into tractable sub-problems.
  3. Balance the inside and outside views. Start from history, then adjust for what is different.
  4. Balance under- and overreacting to evidence.
  5. Distinguish as many degrees of doubt as the problem permits.
  6. Look for the errors behind your mistakes, without falling for hindsight bias.

How to become a superforecaster

You can practise every one of these habits:

  1. Forecast in public. Platforms such as Good Judgment Open and Metaculus pose real questions and score every answer.
  2. Always write a number, with a date and one sentence on why.
  3. Start from a base rate before you read the news.
  4. Score yourself. Record your forecasts and outcomes, then check your Brier score and calibration. Our free Brier score calculator does both.
  5. Compare with the crowd. A prediction market price is a strong benchmark; see how to research a prediction market.

Superforecasters vs. prediction markets

Both pool information, in different ways: superforecasters by careful judgment, often in teams, and markets by letting people bet on their beliefs. Each can beat the other on different kinds of questions. Markets struggle with long shots and thin trading (how accurate are prediction markets?), while a small group of forecasters can miss news a market absorbs in minutes.

Superforecasters vs. AI

The superforecasters' habits are concrete enough to build into software, and in 2026 the best AI forecasting systems are close to them: statistically tied on one major benchmark and narrowly behind on another. We cover the evidence in AI vs. superforecasters and how such systems work in what is an AI superforecaster?

Sikt Intelligence is building an AI superforecaster: it reads the news and data that exist today, checks every source, and independent AI forecasters agree on one honest probability. Every forecast is graded, misses included.

A live AI forecast, scored in public. In the Sikt Midterm Bench, Sikt forecasts the 2026 midterms next to Kalshi and Polymarket. Latest round, Oct 3, 2026: Senate: Democrats 56% (markets 64% to 66%), every round; House: Democrats 88% (markets 91% to 92%), every round. Every forecast will be graded after election day, misses included.

Key takeaways

  • Superforecasters are the most accurate forecasters, identified in a four-year US intelligence tournament (2011–2015).
  • The Good Judgment Project's forecasters beat other research teams by 35–72% and intelligence analysts with classified information by more than 30%.
  • Their edge is method: probabilities, base rates, small updates, open minds and keeping score.
  • Anyone can practise these habits on public forecasting platforms and by scoring themselves.
  • The best AI forecasters are now close to superforecaster level on some benchmarks.

FAQ

What is a superforecaster?

A forecaster whose predictions of world events are consistently more accurate than others', measured with proper scoring rules over many questions. The term comes from the Good Judgment Project, which won a US intelligence forecasting tournament from 2011 to 2015.

How do you become a superforecaster?

Forecast many real questions in public, always as probabilities, start from base rates, update in small steps, and score yourself with the Brier score and a calibration check. Good Judgment Open and Metaculus are good places to start.

Are superforecasters better than experts?

On many geopolitical questions, yes: Good Judgment's superforecasters were more than 30% more accurate than intelligence analysts with access to classified information.

Is AI as good as superforecasters?

Close. In 2026 the best AI systems are statistically tied with superforecasters on ForecastBench, while top human forecasters still lead narrowly on Metaculus.

Sources

Keep reading