Decentralized insurance using prediction markets and game theory

It sounds silly, but they only work on things that are in fact predictable:

Probability of a bike being stolen: Okay.

Major conflict that kills more than 10 million people in the next 10 years: [Might as well read tea leaves](https://www.edge.org/conversation/nassim_nicholas_taleb-the-...)

The problem of figuring out the probabilities in the first scenario is so tractable that the use of a prediction market becomes almost pointless. The only advantage the market has over a data processing algorithm (or even just [bob](http://intheoreum.org/#whatis)) is that it can react quicker to changes in the external world, say when a market participant goes out stealing bikes...