Where’s that guy who was giving it the biggun about his educated understanding of betting markets?
It's funny. He was talking about neural nets. As a fellow data scientist with a masters in machine learning (and I'm sure more qualified PhD folks like @Revan will agree) - no matter how good your model is, it's limited by the data you feed it. And there's just too many variables when it comes to stocks and elections, and you will always have a considerable margin for error because your feature set cannot simply include all the chaos of real life. With a carefully curated data set, an intuitively selected and constructed ensemble and a skillful tweaking of loss functions and hyperparameters, you can outperform other models, - but you can't become fecking Nostradamus.