The implied probability of Brexit a few days before was around 25%. The implied probability of Trump beating Clinton a few days before was around 20%.
Neither the time value of money nor the cost of trading explain these dislocations.
Neither the time value of money nor the cost of trading explain these dislocations.
It was one of the 25 days out of 100 not one of the 75 days out of 100.
> these dislocations
The market puts the chance of flipping a coin and getting heads at 50%, when it comes up tails it doesn't mean that the market was irrationally pricing heads.
The real way to measure predictive accuracy is to find _all_ the times they gave something a 25% chance of happening. If the predictions are accurate then roughly 25% of those things should have happened.
What you should optimise instead is something like log loss between the given probability and the true outcome (0 or 1). That way you’re rewarded not only for being right, but for being confident and right.
Meanwhile, both polls and analysts got those elections completely wrong, much more so than prediction markets. (IIRC, 538 was the only outlet that gave Trump a >20% chance to win, and they got a lot of bad words for their prediction pre-election.)
Unfortunately, people keep falling for this. You can always cherrypick elections from 7 years ago as "evidence" against prediction markets, but if you look at the bigger picture, you will see that they get a vast majority of elections right. But when they don't, even just once in a decade, they get all the blame.
There's plenty of research on this: https://researchdmr.com/RothschildPOQ2009.pdf
Honestly, why hasn't probability-centric mathematical pedagogy taken the world by storm?
Unclear if you think this is incorrect?
Obviously it happened, but most reasonable prediction methodologies put it between 20 and 30%. Notably 538 and Superforecasters (via The Good Judgement Project at the time) both had it in that range (as did prediction markets). There were some outliers - in particular some newspaper(?) had a statistical methodology that gave Clinton a 95%+ chance. In a two horse race that is obviously wrong!
I was working in the forecasting field at the time, and that seemed about right. It was an unlikely but not impossible outcome.