As mentioned on the page about how Ken Green became the world’s richest man, the key was getting reliable predictions of future events from several people by minimizing their error covariance. How to do that?
To find suitable people to do predictions is itself is a non-trivial problem. Finding groups of people with low error covariance is much more difficult. To do these tasks requires massive amounts of data about large numbers of people. It is actually not hard to find out almost everything about everybody by using a technique called Recursive Exhaustion.
The book does not comment on this method and actually describes much more conventional and less effective methods, but something like it seems essential.