Hacker News Reader

They can pick up a lot of signals, and they have claims data: it seems like they should be able to build a pretty good model.

Even just "number of miles driven" is a strong signal.

Progressive also logged the time of day you drive. Between 12-3am (IIRC) was marked the highest risk time to drive.

Might actually be pretty interesting to see how the random statistics collected correlate with claims. It might indeed turn out fast, aggressive drivers never file claims, so ... that's what should be rewarded.

Sort of the inverse of the self driving car problem - they follow all of the rules, all the time, and are more accident prone, even though they're not at fault for the accident.

http://time.com/4098303/self-driving-cars-accident/

Still probably less crude than your age and gender, no?

Those methods have a lot more refinement, but eventually, most certainly.

I suppose the more important argument is not about "what's the quality of the model?" vs. "what goal is the insurance company trying to accomplish with better risk pricing?"