I have done prediction to death (see here) The patron saint of pointing out that experts don’t have clue is Philip Tetlock.
This is an extended piece and video with him. It takes a bit to get through but the central message is that models beat experts as evidenced in the recent US election.
The interesting thing about this movement away from experts and towards models is that it is now being taken seriously and is no longer a fringe movement. For example Kaggle attempts to match modellers and data miners with quests in a competition format. A quick look at their home page shows that some folks are very serious about this with the likes of GE offering a prize of $500k for someone who can mine the data they provide to generate an algorithm for real time flight control.
For those who still watch the rubbish that passes for financial commentary on the Fox network consider the following –
1. The more self confident an expert appears the worse their track record is.
2. Forecasters who get a single big outlier correct are more likely to underperform in the future. The best example here is John Paulson – fluked the housing crisis but has absolutely toasted his hedge fund in the years since.
3. Expert forecasters actually do no better than an audience poll. This means that experts ad not value or alpha to the information they present. You might as well ask the bloke serving coffee in the cafeteria.
