Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

This election proved that virtually the entirety of the election prediction space has no idea what the hell it's doing.

We live in a world of big data with vast computational resources, and this outcome is somehow an upset—a totally unexpected surprise.



Even modern computing is still vulnerable to the old GIGO phenomenon: Garbage In, Garbage Out.

The way data was collected has some very serious flaws.


538 gave more than 25% chance to Trump winning. That's losing two coin flips, not "totally unexpected".


And in fairness to Nate, he was saying the polls were about 42 Trump 45 Clinton with ~13% undecided and those undecideds could send the election either way. With hindsight I'd say quite a few undecideds were Trump supporters who were afraid to admit it which I can understand. Just look at the reaction to Theil backing Trump with various calls for him to be fired. I can see why people kept quiet about it.


The degree by which Trump won is the "totally unexpected" part.


Only just? He even lost the popular vote.


I wonder if facebook and google could've built a better model. Prediction based on polling doesn't seem as reliable compared to what people post to their newsfeed or search for.


A major problem is that the electoral system is chaotic. In the sense of chaos theory: small deviations in input (votes) lead to large differences in the result (electoral college votes). The pollsters were off only be like 2-3% or so, and a bit the distribution.

Also, it appears Trump Supporters were often too ashamed to be honest to pollsters.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: