Superforecasting

This is the sequel to Tetlock's earlier finding that expert prediction is barely better than chance, and it pursues the part of that result which was under-reported: some people really are good at it. The evidence comes from a large government-funded forecasting tournament, the Good Judgment Project, in which tens of thousands of ordinary volunteers forecast world events. A minority did consistently well, beating benchmarks, competitors and prediction markets, and in some comparisons beating intelligence analysts with access to classified material. The book's subject is what those people do differently, and the answer is method rather than genius. They break a question into parts, start from how often the thing happens in general before adjusting for the specifics, think in degrees of probability rather than in yes and no, update in small increments as evidence arrives, and keep score so they can find out when they were wrong. Tetlock's claim is that this can be taught and measurably improves accuracy. It is among the better-evidenced arguments in the field.

Written from the publisher description, Open Library.

What this vault takes from it

Reliability

Not audited here. 1 note in this vault draws on it.

People

• Philip Tetlock