Days before a referendum that seemed to be in the bag for British Prime Minister David Cameron, a YouGov poll suddenly showed the "Leave" side ahead.
The pound plunged, the Prime Minister panicked - then Scotland voted to stay in the United Kingdom by a 10-point margin.
Two years on, and British voters are preparing to decide whether to leave the European Union on June 23.
The question for politicians and pollsters studying the current referendum is whether recent polls showing "Leave" in the lead, including ones from YouGov and TNS, are just a symptom of another wobbly weekend before a clear Cameron victory or a sign that his luck has run out.
Pollsters aren't much help. Having been burned in last year's UK general election, when they failed to predict that Cameron's Conservatives would win a majority, many have doubts about the reliability of their work on the referendum.
The pound fell last week when ICM published online and telephone polls both showing "Leave" ahead. It fell further yesterday after YouGov put "Leave" at 45 per cent to 41 per cent for "Remain" and TNS had the Brexit camp ahead by 43 per cent to 41 per cent.
"It's hard to think of a more bewildering electoral event," said Martin Boon, a director at ICM. "The polls have not really moved, and more phone polls of late with their pro-'Remain' tendencies have added to or created the narrative that 'Remain' might cruise it. But that could be a false narrative, and for me the only correct thing to continue to say is: I just don't know how this will go."
A year-long inquiry into the May 2015 UK election polling failure concluded that the main problem was unrepresentative samples, and that this would be hard to fix.
Instead, the inquiry urged the public to be more sceptical about polling. If more reason were needed to doubt the EU referendum polls, they don't even agree on a picture. Online polls have tended to show the two sides tied, while telephone ones have generally put "Remain" ahead.
It's not just pollsters who are anxious. People within the "Remain" campaign, speaking on condition of anonymity, said they thought those who expected a clear win for their side were being far too complacent. In particular, they had doubts about whether their supporters could be relied on to turn out and vote - posing another problem for polling.
"There's a discrepancy in levels of motivation," said Joe Twyman, head of political polling at YouGov. "There are millions of people who would walk barefoot across broken glass to vote to leave. The 'Remain' campaign doesn't have people who feel the same way. I still think we'll vote to stay, but that's assuming that there's a move towards the status quo in the final days."
The "Remain" campaign's problem is that it needs young people to vote - and they are the ones least likely to do so, according to Twyman. He said they needed endorsements from young celebrities, such as the junior members of the royal family. "If I were the 'Remain' campaign, what I'd want is for William and Harry to come out in support, perhaps in an accidentally-on-purpose open-mic moment."
Those running the "Remain" campaign aren't simply looking at the headline numbers. They're more interested in polling that shows voter attitudes, and here, too, there's cause for concern.
The pro-EU camp's core message is that a Brexit will make the country poorer. Yet 58 per cent of respondents to an Ipsos Mori poll published last week said they didn't think leaving the EU would affect their own standard of living - indicating the Government's message has yet to get through.
But some pollsters are less uncertain. Tom Mludzinski, director of political polling at ComRes, is more confident of a "Remain" win.
"'Leave' have the hard work to do to convince and persuade voters," while the pro-EU side "have to motivate voters," he said, noting that polls have traditionally understated the status quo option of staying in the EU, and phone polls are largely showing comfortable leads for staying in. "For a high chance of Brexit you'd want to see a consistent lead for 'Leave,' which they haven't yet had."
Tywman put the probability of a Brexit at about 33 per cent. Betting markets are fast catching up with that, according to Oddschecker, with the implied probability of a Brexit rising to 30.9 per cent, up from a low of 19.7 per cent on May 26. Matt Singh, polling analyst at NumberCruncherPolitics, had the probability at 21.7 per cent.
"There are some signs of swing towards 'Leave,' but the evidence is patchy and not entirely consistent," Singh said. "Even though 'Leave' may have caught up a bit, it's still behind and time is running out."
Damian Lyons Lowe, chief executive of Survation, was sceptical about polls showing 'Leave' ahead. He suggested people who took part in online polls might be more engaged with the news, and therefore affected by the media narratives, than the ordinary public.
"The movement in the polls might be right, but the lead we should be wary of," he said of the 'Leave' campaign. "Over the next couple of weeks, worry about Brexit will increase, volatility will increase, but the probability of things happening that will help them will drop."