David Cameron and the Conservatives came out on top of the 2015 general election – alongside a hitherto unknown blogger called Matt Singh.
The former Barclays interest rate trader had founded the Number Cruncher Politics website in 2014, and just a year later he would become a minor celebrity in Westminster.
Singh, unlike Britain's major pollsters who had Labour and the Tories neck-and-neck, predicted that the Conservatives would win a slim majority in the House of Commons.
With another general election scheduled for 8 June, Singh, who will release his first prediction after the local elections on 4 May, sat down with IBTimes UK to discuss the vote and whether the electorate can trust the polls again.
Singh: People are absolutely right to be a bit sceptical about polls. The first election I remember was 1992 [when John Major secured a surprise majority for the Conservatives] so I've always been somewhat of a sceptic myself. I believe there are two extremes that people can go to and they're both wrong.
One is the sort of thing that we saw before the last election where people were looking at polls to a decimal place and thinking that it is absolutely right and the Conservatives can't win a majority.
The other extreme is where people don't trust them at all and just start making up assumptions. I think healthy skepticism, but also an acknowledgement that, although they're not perfect, they are the best guide we have.
IBTimes UK: What went wrong in 2015?
Singh: There are quite a few ways where polls can go wrong. The particular thing that happened in 2015 was that pollsters were not succeeding in getting a representative sample of the population.
You don't need to speak to everyone, you can get a reasonable margin of error if you speak to 1,000 or 2,000 people.
But those people need to be representative of the population as a whole and they were struggling to do that in that case – they were getting too many people who were interested in politics...and that messed up the flows of voters between parties that they recorded, and it ended up throwing the results off.
IBTimes UK: Will the Conservatives' high poll ratings impact on the party's turnout?
Singh: In terms of the picture of the polls and how that might affect turnout and people's assumptions about the result, it is a very similar situation to what we had two decades ago when Tony Blair and Labour were widely expected to win by a landslide [at the 1997 general election].
They were very keen to push home the message, as the Conservatives are now, about not being complacent and not trusting the polls.
Because, then as now, the polls had got the previous election horrendously wrong. On that occasion it worked. Whether it will work this time remains to be seen.
Any time you get a predicted landslide it does tend to depress turnout to some extent. So it will be interesting to see if that happens and how it happens. It depends who stays home, which is really what is going to determine the election.
IBTimes UK: Will Brexit dominate the general election as an issue?
Singh: It's interesting because quite often elections get framed about something and certainly that's how it's reported and that's what politicians want to talk about.
In the end it's ultimately up to the voters and there haven't been very many polls specifically on that question done since the election was announced.
The two things historically that have been the strongest predictors of how people vote are their perceptions of the leaders and their views on which party would be best on the economy.
At the moment both of those are saying pretty much the same thing - they very much favour the Conservatives.
IBTimes UK: How should people new to politics read the polls?
Singh: The biggest mistake that newcomers often make is putting too much emphasis on the poll that looks interesting.
That is a problem, because people will often tend to talk about the ones that look interesting rather than the ones that look the same as the others...The important thing is that if you don't know anything about the individual polls look at averages. If you look at averages, look at the trends.