In today's age we have some of the most accurate and in depth polling that has ever existed, It can predict with a reasonable degree of certainly of how an election will go, sometimes if the race is only separated by a small margin then the election can go either way regardless of who has the slim lead in the polls.
Election polling data can, and has been used to predict elections in the past. They are scientifically valid and proven. They include reasonable margins of error, and can generally provide a working knowledge of the American public in a way that is predictive. Although not perfect, it is quire useful to use.
In the majority of our recent elections, the popular vote has not been matched equally with the electoral outcome meaning that even if people were polled in the most efficient way possible, the Electoral College still trumps opinion and the desires of the American People. While the Electoral College is supposed to reflect the opinions of their specific region, we have seen more and more of popular votes (votes actually made by the people) going to one candidate while the electoral votes (votes made by the Electoral College) lean more to the other. If the American People, their opinions, and their desires are supposed to select their Commander in Chief, then why is the popular vote, or opinion polls period, not relied on more or even used as the deciding factor rather than the Electoral College? As long as another party stands in the way of the collective opinions of the American People, then no amount of pre-election polling can add any accuracy to predictions made about the selection of the President or any other leaders.
The people who participate in these polls always vote on their preferred site. They wouldn't know if a competing site had a poll. So whatever bias the site they visit has tends to skew the numbers in the site's candidate's favor. Not to mention, the parties themselves have their own trends. Democrats seem to be more accepting of technology than Republicans. So they are more likely to join in on a poll.
There are too many factors in polling to rely on them for accurate projections. I suppose they are our best way not to predict future trends in the mindset of people, but so often does public polling fail or fall completely short all together. This was proved in several elections.