4 ms·
> What is the practically relevant difference between 100% prediction capability and 99.999% prediction capability? If there were 100% prediction capability,
by chimeracoder 8y ago
> What is the practically relevant difference between 100% prediction capability and 99.999% prediction capability?
If there were 100% prediction capability, there would be no risk to smooth, and there would be no reason to buy insurance, because the insurance would be more expensive than paying the costs of the would-be insured risk directly.
- zAy0LfpBZLC8mAC 8y agoBut with 99.999%, there is a risk to smooth in practice? How about 99.99999999999999999999999999999999%?
- chimeracoder 8y ago> But with 99.999%, there is a risk to smooth in practice? How about 99.99999999999999999999999999999999%? Uh, I guess this question might be interesting from an academic perspective, but not a practical one. There's no risk out there which can be predicted with 35 nines of precision, let alone one that people are interested in insuring. And certainly not one in which the overhead of providing services wouldn't inherently dwarf any benefits to both parties.
- zAy0LfpBZLC8mAC 8y agoIn other words: There is no practically relevant difference between a prediction ability of 100% and 99.99999999999999999999999999999999%, which is exactly the point I was making: There is is nothing magical about the 100% mark--free will or assuming perfect information is irrelevant to the problem at hand, insurance becomes asymptotically less useful the better risks are modeled and used to set premiums where the premiums cannot work as an incentive to avoid the risks in the first place.