They're always wrong.

Or you might as well get monkeys to chuck darts at a board.

That's not quite fair.

They don't completely useless.

Well, you do have to be careful how you interpret them.

So here's some things to bear in mind about how polls work.

First, you can't ask everyone.

More than 45 million people can vote at the next election.

You really can't talk to the rule.

So pollsters usually asked between 1000 and 2000 people a simple question.

If there were general election tomorrow, which party would you vote for?

But if you only asked pensioners or people in Wales or the crowd of the Labour Party conference, this wouldn't be representative of the UK as a whole, and you get a biased results.

So polling companies try to build a carefully balanced sampled to get a snapshot the national mood from just a couple of 1000 people.

But even with a representative sample, there are other factors to consider.

For example, older people tend to be more likely to vote for younger people, and pollsters have to factor this into their calculations.

Turnout among different groups of voters sort of the hardest things pollsters to predict accurately.

Second, polls aren't perfect in the recent past.

They've got it wrong because their samples weren't quite right.

In 2015 pollster counted too many Labour supporters and not enough of the people who voted to keep David Cameron in Downing Street.

But two years later, the polls were wrong the other way and failed to pick up the scale of the Corbyn surge.

They're often big differences between pollsters, and they can't all be right.

So a bad sample commune a poll misses the mark.

Bathing with a good sample.

Polls cannot pinpoint a party's exact vote share.

That's because of something called the margin of error when pollsters say their party's on 40%.

What they really mean is that confident the party is somewhere between 36% and 44% so neither party is up a bit in one poll.

This may not be a genuine shift in the public mood.

Third, despite all this pulls often are quite accurate.

For example, they did a pretty good job in recent European Parliament elections in the UK, where polls have got it wrong that generally haven't been off by too much.

We've had lots of close elections recently.

A winner racist tight, a small polling mixed mean getting the winner wrong.

Fourth vote share doesn't equal.

M P's constituency based voting system in the UK makes it especially difficult to translate polls on National Party support into parliamentary seats.

The Conservatives gained votes in 2017 but lost sees.

That's because it's not just about whether apart it is, well overall about whether they do better than their opponents in key constituencies.

Ignore anyone who has a simple formula to predict who will win the most seats.

So how should you read the polls?

We'll keep in mind.

This is really important.

The polls tell you about people's views a specific point in time.

They're not a prediction of the final result.

So ABC a poll a few weeks away from an election.

The final result may end up very different, not because the poll was wrong, but because people change their minds.

Also, be very wary of any single pole it's best to look at overall trends from several different polling companies are BBC poll tracker does this?