For overmuch of the 2024 US statesmanlike campaign, polls and pundits rated the contention excessively adjacent to call.
Then Donald Trump delivered a commanding triumph implicit Kamala Harris, winning astatine slightest 5 battleground states, and performing unexpectedly good successful different places.
He is present poised to go the archetypal Republican successful 2 decades to triumph the fashionable vote, and could participate bureau with a Republican-controlled House and Senate astatine his back.
So were the polls incorrect astir it being a choky contest?
At the nationalist level, they surely appeared to underestimate Trump for the 3rd predetermination successful a row.
But successful the battleground states specified arsenic Pennsylvania, wherever astir of the polling was focused, Trump's borderline of triumph was typically wrong striking region of his show successful the polls, adjacent if the forecasts were somewhat little than the last outcome.
The mean polling mistake seen successful these plaything states was not really that big. Still, successful choky campaigns, tiny changes tin marque a large difference.
Ahead of the results, media outlets specified arsenic the New York Times warned that contempt the adjacent contention successful the polls, it could extremity up looking similar a landslide for either candidate, fixed the borderline of error.
In immoderate little closely-watched parts of the US, polls underestimated Trump's enactment much importantly - a motion of immoderate unsighted spots, said Michael Bailey, a prof astatine Georgetown University and writer of the publication Polling astatine a Crossroads.
"At a glance, successful the battleground states, polls ran a small blistery for Harris but truly not truthful bad, but erstwhile you excavation deeper, it's each a small little impressive," helium said.
In Florida, for example, polls tracked by RealClearPolitics successful the last weeks of the predetermination enactment Trump up by astir 5 points. He won by a greater borderline of 13.
In New Jersey, Harris was expected to triumph by astir 20 percent points, based connected the 2 astir caller polls tracked by the site. Her borderline was much slender, person to 10.
"Just ideate if we knew that oregon had a amended consciousness of that a period ago. I don't cognize that it would alteration the predetermination but it would surely alteration our expectations," Prof Bailey said.
He said pollsters this twelvemonth whitethorn person leaned excessively heavy connected bets that radical would behave astir arsenic they did successful 2020 - failing to expect the extent of the plaything among Latino and young voters toward Trump.
"These models that presume truthful heavy that we're going to get a repeat... they're a catastrophe erstwhile there's a large change," helium said.
But pollster Nate Silver, laminitis of the 538 polling investigation site, said determination were signs of peril for those who were looking.
Writing astir the predetermination results, helium noted a canvass from New York City past month, which indicated Trump making large inroads successful the accepted Democratic stronghold.
"This is simply a occupation the enactment should person been prepared for, due to the fact that determination was plentifulness of grounds for it successful polls and predetermination data," helium wrote.
Debate astir the polls is definite to proceed successful the months ahead.
That is peculiarly existent successful a twelvemonth successful which figures similar Trump and billionaire protagonist Elon Musk person promoted betting markets - galore of which did forecast a decisive Trump triumph - arsenic a much close alternative.
Experts said the polling assemblage does look challenges.
Response rates to surveys person plunged, arsenic it becomes easier for radical to surface calls from chartless numbers.
The autumn has besides coincided with rising distrust of media and institutions - a diagnostic peculiarly pronounced among Trump supporters that immoderate reason has led to their under-representation.
Prof Bailey said the large miss successful the much-discussed canvass of Iowa by Ann Selzer - that was released days up of the predetermination and indicated a three-point pb for Harris successful the authorities - showed the risks of the accepted approach.
To marque up for specified issues, galore of the astir high-profile polls present relation much similar models, with firms weighting responses from antithetic groups and making different assumptions astir factors specified arsenic turnout.
Many pollsters person besides shifted to utilizing online surveys, but experts said those were known for being unreliable.
This year, voters who were inclined to capable retired online polls were much apt to beryllium Democrats, James Johnson of London-based polling steadfast JL Partners told the Times of London newspaper. They were "more apt to beryllium young, they’re much apt to beryllium highly-engaged, they’re much apt to beryllium moving from home," helium said.
Prof Bailey said pollsters had to "move on" from random samples and get comfy with modelling, portion doing a amended occupation investigating and explaining their assumptions.
But Prof Jon Krosnick of Stanford University said without genuinely random samples, surveys would stay susceptible to error.
Pollsters are "trying hard but they support trying to beryllium excessively clever," helium said. "What we request to bash is spell backmost to basics and walk the clip and wealth that it takes to bash polls accurately."
North America analogous Anthony Zurcher makes consciousness of the contention for the White House successful his twice-weekly US Election Unspun newsletter. Readers successful the UK tin sign up here. Those extracurricular the UK tin sign up here.