The only thing certain about opinion polls is their uncertainty

I try to remain sceptical at all times and prefer to use opinion-poll averages as a way of screening out outliers and focusing on trends

Related: General election 2019 – what you need to know (Getty)

Every election, we journalists resolve not to allow coverage to be dominated by the opinion polls. Every time, we find ourselves fighting a losing battle against the human desire to ask who is going to win, and how much by?

Last week there was too much attention given to the YouGov MRP poll, published on Wednesday night, which was treated as holy writ because a poll using the same method was unexpectedly accurate at the last election.

I am guilty of contributing to the hype: I wrote about it, because I am interested in polls. I think they are the worst way of knowing what people think – apart from any other method.

But I hope I resist the temptation to assume that they are infallible, which seems to be just as wrong as dismissing them altogether. I try to hold the uncertainty that surrounds them in my mind at all times. Martin Boon, formerly of ICM and now of Deltapoll, said last week that opinion pollsters are much more sceptical about their product than most of their consumers are.

I try to emulate that scepticism, but there is always a pressure in journalism to simplify, which means a tendency to treat a set of figures as definitive, and to sensationalise, which means paying more attention to a surprising figure than to one that shows no change.

That is why I prefer to use opinion-poll averages, as a way of screening out outliers and focusing on trends. And that is why, whenever I can, I try to give the range of likely outcomes suggested by a poll.

The YouGov MRP poll, for example, was headlined as suggesting a Conservative majority of 68 seats. But the poll actually suggested a possible range of outcomes, from a Tory majority of six to one of 120. That means that if the sample design is good, there is a 95 per cent chance that the true picture lies between those two numbers – although a number in the middle is more likely to be right than one at either extreme.

Humans are not good at dealing with uncertainty, but it is worth remembering, when anyone makes a confident prediction of the election outcome, that this probably reflects a wide range of possibilities. Currently this is tilted towards a Tory majority of some kind, but a hung parliament remains a possibility.

Yours,

John Rentoul

Chief political commentator