When people are asked who they will vote for in an upcoming general election they tend to tell you what is their hearts desire, if they've made up their mind at all. However peoples' voting habits are not quite clearly as straight forward. On the day voters apparently change their minds rather swiftly and will vote tactically.
For constituencies that are typically a two horse race, people tend to vote for the party that will keep the other out. For example, Warrington North shows a rise in UKIP support. Polls show UKIP in a tight second above Labour, with the Tories in the lead. On the day however, should it be shown that Warrington North is in a close race between the Conservatives and Labour, most of those UKIP supporters will quite likely change their vote to ensure the party they dislike the most doesn't win. It's a necessary evil.
This is why the polled UKIP support, which is considerable, may not translate into any more than a couple of seats and will most likely not even eclipse the Liberal Democrats. Almost certainly not the SNP.
So why would polling for the SNP be more accurate? Well put simply, the SNP are the only viable alternative party to Labour. There are no other parties with a realistic chance for a win. Even the Lib Dems are unlikely to be seen as a viable third party. Unlike UKIP who are expected to take a few seats on the coastal towns of south east England, the projections for SNP seats is so great that it makes them an easier party to vote for in terms of a win. For some reason, people like to back the winners, so long as their policies are somewhat aligned to their politics, else they consider their vote wasted. UKIP have an image problem with anyone below the age of 50 whereas the SNP doesn't seem to have that issue.
Elsewhere, Liberal Democrats are fighting for their seats and it seems, despite their broken promises and pledges, some of their MP's may well survive, thanks in some part to loyalty. People will likely vote for an incumbent if they are a good MP who will deal with their local issues regardless of the party the MP is affiliated too. A notable example is Devon North. It has traditionally what you'd expect to be Tory voters but are likely to vote back in the Liberal Democrat incumbent because he's perceived to be an excellent MP for the constituency.
So there we have it. Polls tell us a story but they're not as useful as we may be led to believe. In this regard the politicians may be correct. Polls are not that accurate a prediction. What we do know is that polls can galvanise support at the last second. For that consideration, they cannot be disregarded completely.