Sadly though referendums are very difficult to pass due to requiring a super majority or whatever, the majority of the majority of people in states. The polls are pretty grim, I don't think it's likely at all.

There should be something like Heisenbergs Uncertainty Principle for politics (and movie reviews). By observing, analysing and publishing, pundits will (and are probably trying to) affect the outcome.

There is actually something to that. Statisticians look for correlations in voting trends to help predict future results. For instance, some analyst finds that one suburb in Michigan strongly represents the results of every election for the past 50 years. So, the news media picks up on the story and uses that to talk about how the candidates stack up in that suburb. This makes the candidates focus heavily on that suburb, so that the news will say their campaign is strong, which makes people think they are a good candidate, so they get more votes. But candidates could be hyperfocused on the one suburb and don't bother polling the surrounding counties where they have fallen far behind. All it takes is for some people who haven't traditionally voted in large numbers to show up at the polls, and the campaign is caught sleeping.

According to the Experts

I don't recall "the experts" guaranteeing any of the things you listed would happen.

When I checked the weather report this morning, it said 20% chance of rain. It hasn't rained yet, but it might later. But rain or no rain the weather report was accurate.

If it was possible to predict the outcome of an election, there wouldn't be any point having an election at all. The other candidates would save themselves all the money and effort by just not competing at all. Similarly, if the yes/no vote had an obvious winner it would have unilateral support by both major parties.