Superforecasting: The Art and Science of Prediction

by Philip E. Tetlock

If we are serious about measuring and improving, this won’t do. Forecasts must have clearly defined terms and timelines. They must use numbers. And one more thing is essential: we must have lots of forecasts.