Forecasting and the wisdom of crowds
We cannot predict the probability of future events with certainty, but we can attempt to do the best possible job.
It has often been observed that large groups of forecasters collectively perform better than individuals - the so-called 'wisdom of crowds'. But this isn't just a matter of taking the mean prediction. There is more wisdom in the distribution of people's forecasts than simply the average.
We try to extract this wisdom. For example, with US political scientist Philip Tetlock and INSEAD data scientiest Ville Satopaa, we created the 'skew-adjusted extremized mean', which identifies small groups of especially well-informed forecasters. Another example, created with our student Zane Hassoun, is 'kairosis', a means of aggregating forecasts over time by using Bayesian techniques to identify moments of change in the distribution of forecasts.
An informal introduction to the basics of probability scoring and forecast verification, inspired by the Department of Mathematics' FIFA World Cup prediction competition, was the cover feature in the May 2025 issue of Significance, the joint US-UK magazine for professional statisticians [1].