Hacker News Reader

Further, they should be telling people a model is only as good as the data it has as input. If the data only sees upward movement in valuations then it will only predict upwards. A model can’t tell you anything about a situation where the underlying factors are different from its data. Same thing happened with mortgage backed securities in 2007/2008. The math was very complex and still couldn’t see around a corner.

View on HN

The bottom line is that the domain here is under the umbrella of the social sciences. The phenomena the models describe is historical, and how closely they will match what happens from that point on relies on social stability. In the grand scheme of things, social stability is ephemeral. It isn't rocket science.

I'm not making a joke. Newton published his models in the late 17th century. The models hold no better or worse today than they did then. If you had rockets in 1687, they would work the same way then that they do now. But, social science is nothing like hard science. I worry on some fundamental level, even people who will claim that they "of course" know this in fact forget this. There is hubris here.

Reply on news.ycombinator.com