Intrade had odds of the Supreme Court overturning Obamacare at 96%. What's your point? 
There's a big difference there. They had to guess what 1 or 2 members of the SCOTUS would do.
In this case, they have to guess what 100 million people will do. MUCH easier.
Surely you can tell the difference here? It's much easier to predict things with bigger numbers and ANY kind of sampling size. They couldn't exactly survey the 9 people on the court.
A single decision of 1 man is WAY tougher to predict, than the decisions of 100 mil people who are being polled.
When it comes to presidential elections, INTRADE = more accurate than Gallup, Rass, 538, etc. No arguing it.