French voters did exactly what they were expected to do -- to the relief of French pollsters who helped redeem a profession still smarting over getting Brexit and the US election wrong.
The final surveys published on the eve of Sunday's presidential election showed pro-EU centrist Emmanuel Macron and far-right candidate Marine Le Pen qualifying for a run-off vote in two weeks' time.
Macron was projected at 23-24 percent with Le Pen at 22-23 percent -- and those figures were in line with their final scores of 24.01 percent and 21.30 percent.
It was the same success -- forecasts that were within around one percentage point of their final results -- for scandal-hit conservative Francois Fillon and far-left firebrand Jean-Luc Melenchon.
"There's a feeling of relief in the sense that polling groups have been under attack a lot during the campaign," pollster Yves-Marie Cann from the Elabe group admitted to AFP.
Britain's unforeseen vote to leave the European Union and Donald Trump's shock victory in the 2016 US election "spread suspicion over the whole profession" despite the generally good record of French pollsters, he added.
The French campaign was dogged by constant speculation, with some pundits raising the spectre of a "hidden vote" for rightwing Fillon or Le Pen, sampling errors or a herd mentality among the pollsters.
Only once have French presidential pollsters failed badly in recent memory, Cann said: in 2002, when far-right leader Jean-Marie Le Pen, father of Marine, stunned France and the world by making it through to the second round.
In the days ahead of Sunday's vote, "they successfully detected the late rise in turnout and indicated Emmanuel Macron in the lead and Marine Le Pen in second," said Anne Jadot, a political scientist at the University of Lorraine in eastern France.
- High turnout helps -
A combination of factors probably explains the French success after perceived American and British failures.
French polls are constructed differently, with pollsters typically seeking out a sample of people who are chosen as being representative of the voting population at large.
US and British pollsters often contact a random sample of voters by phone or internet.
Turnout for French presidential elections is generally high at around 80 percent, making voters easier to second-guess. High levels of abstentionism make an electorate more difficult to model.
And French presidential elections are a simple process, with the winners in the first round being the two candidates who gather the most votes nationwide.
Macron and Le Pen will go through to a run-off on May 7 when one of them must get more than 50 percent, with Macron widely predicted to be the eventual victor.
The American presidential system is indirect and far more complicated. Presidents are elected on the basis of votes cast state-by-state with the winner needing at least 270 votes from an electoral college.
Some political scientists believe American pollsters have been unduly criticised.
They successfully forecast the popular vote showing Democrat Hillary Clinton beating Trump by 48 percent to 46 percent, but not his victory in the electoral college.
In Britain, modelling the campaign dynamics of the in-out EU referendum last June posed technical difficulties and the final result of 52 versus 48 percent was within the margin of error of some polls.
But the credibility of the British industry was already in doubt after a general election in 2015 when it failed to detect a surge in support for the Conservative party of then-prime minister David Cameron.