[Platapus]

First of all, I question your assertions that polls are wrong most of the time.

A properly designed and implemented poll should be representative on the average. That means that there WILL be times when the poll is not representative of the specific population

Since polling almost always involves trying to represent future human action, there is added complexity.

The polls may be constructed and implemented incorrectly (either deliberately or through negligence)

The polling population may not be representative of the desired population (think Dewey beats Truman)

People lie on polls if they believe they can influence future events.

People's opinions may change. If a population is polled six months before an election and the outcome of the actual election differs from the polling results, that may not be the fault of the poll.

Polling and voting are emotionally distinct and separate activities. Voting in done in private, polling is done in front of a person. A person may be reluctant to state their actual opinion face to face, but will not have such inhibitions when voting in private.

Polling and Surveys used to be my job back when I was a govie. It is a hard job designing and populating a survey/poll fairly. Not everyone is willing or capable of doing it correctly.

There are many reasons why a poll/survey can end up not accurately representing an actual future event. Predicting the future is hard.

And I have not even touched on the complexities of analyzing and generating the statistical results. That's a whole different matter.

Personally, I only look at the first number in the results. +/- 10 percent is, to me, probably even. Only if there is a 20-30% difference do I start to consider that there may actually be a difference.

I shake my head ruefully when the news media is comparing decimal percentages as being significant. Yikes!