Don’t worry, Lore can still be reassembled.
To be deliberately contrarian:
Sexist.
Racist
Heterophobic.
Racist again.
Again, being deliberately contrarian here.
People forget that polls were pretty wrong last election too, but in Obama’s favor. Nobody cares when the numbers are wrong as long as the correct person wins.
The other important facet of a statistical survey that is almost never reported in the press is the Confidence Level (CL). While margin of error (MoE) is exactly as dracoi explained, CL says “the inverse of this number is the frequency that this survey will be completely wrong”. Given the general MoE range of 3% and the pool size of 1,500 on the polls I saw, the CL would have been 95%. A CL of 95% means that 5% of the time the poll will be so far off as to be useless. Five percent is one in twenty, not exactly rare.
Another piece the public never gets is the ‘story telling’ aspect. Numbers tell a story, but not always an obvious one. A large part of an analyst’s job is to interpret the numbers and tell the story. I heard one prominent analyst say that his polls showed Trump winning, but he thought the numbers represented a “normal” election. He saw Trump as an “abnormal” candidate and adjusted his numbers downward. Others ‘normalized’ their poll results based on the assumption of a larger than normal voter turnout. They read the story wrong, but that is what was released to the press.
If I roll a 20 at the D&D table, a 5% chance, that doesn’t mean data is dead.