Superforecasting

Philip E. Tetlock and Dan Gardner

The book behind the Superforecasters concept and a key source for Reference Class Forecasting. Reports the findings of the Good Judgment Project, which won IARPA’s 2011-2015 forecasting tournament.

Key Ideas

• The outside view first: superforecasters start from a base rate (“comparison class”) before weighing case-specific detail. This is Reference Class Forecasting under a different name.
• Update in small steps as evidence arrives, rather than holding fixed or overcorrecting.
• Forecasts must be specific and time-bound so they can be scored. The Brier score makes accuracy measurable and forces calibration.
• Builds on Tetlock’s earlier Expert Political Judgment, which found average expert prediction barely beat chance, and the “fox vs hedgehog” split: foxes who hold many partial models forecast better than hedgehogs with one big theory.

Highlights

My Take

See also

• Superforecasters — the concept stub
• Reference Class Forecasting — the method at the core of how superforecasters reason
• Decision Frameworks — where forecasting sits among debiasing tools

Reading log

• 2026-06-21 — added to queue