This book makes a few good points, sandwiched between a lot of elitism. The points I liked were

We are biologically predisposed to reproduction, and not really good at anything else, especially estimating and working with randomness.
The 'old' story of send 50 people a letter saying GOOG will go down, and 50 people a letter saying GOOD will go up in the next week. Once that week has passed, take the 50 people you were correct about, and do the same thing for each half of that group. With each iteration you have a group where you've been right 100% of the time for how every many weeks you want. This 'proven track record' might sound convincing yet completely random.